Managers beyond borders: side-by-side management in mutual funds and. pension funds.

Transcription

1 Managers beyond borders: side-by-side managemen in muual funds and pension funds. Absrac The simulaneous managemen (also called side-by-side -SBS- managemen) in muual and pension funds is increasing; however, i has no been analysed previously. This paper sudies his pracice from boh firms and managers perspecives in UK equiy muual and pension funds. We find ha managemen firms develop hese arrangemens o increase he firm size and raise benefis via fees, raher han o improve performance. Conrary o our expecaions, SBS arrangemens do no pursue o reain alen, so SBS managers presen dissimilar skills: op (boom) SBS managers under- (ou-) perform op (boom) non-sbs managers. We furher examine his resul by analysing heir fund holdings and find ha holding more similar porfolios enhances performance. Accordingly, op SBS and op non-sbs managers presen more similar porfolios han boom managers; however, op non-sbs managers presen beer seleciviy skills. On he oher hand, boom SBS managers ouperform heir peers because he fund similariies do no influence he performance of boom non-sbs manager funds. Keywords: firm, manager, muual fund, pension fund, side-by-side managemen. JEL classificaion: G10, G23 1

2 1. Inroducion. Muual funds and pension funds have become he mos imporan collecive invesmen insrumens wih more han 38 rillion and 24.5 rillion of global asses under managemen in 2016, respecively (INVERCO, 2017). While muual funds are he main channel for reail invesors o paricipae in capial markes (Goyal and Wahal, 2008), pension funds are he main invesmen vehicle for reiremen. In his conex, a number of managemen firms realised he expansion opporuniy of hese wo indusries and decided o marke simulaneously muual and pension funds, appearing he side-byside managemen (SBS hereafer). This pracice consiss in he simulaneous managemen of wo invesmen vehicles by a managemen firm (Cici e al., 2010) and, someimes, by a manager 1 (Nohel e al., 2010). Noneheless, no prior sudies have sudied his pracice in pension and muual funds, as far as we know. This paper inroduces novel discussion abou how hese simulaneous managemen srucures may provide poenial synergies (efficiency of scales and deploymen of alen), or alernaively, be a disracion o he managemen of one vehicle or he oher wheher he SBS arrangemen is exended a a manager level. Financial lieraure usually focuses on funds and/or firms as he analysis unis, we hen also evaluae he role of hese agens and discern wheher managers handling muliple funds in one invesmen indusry behave differenly han managers handling muliple funds in wo invesmen indusries. SBS arrangemens are carried ou for he aracive poenial on boh demand and supply sides. From he demand side poin of view, invesors may receive specialized advice o design specific invesmen plans by invesing in several invesmen vehicles of he same firm (according o invesor s needs and he knowledge of he managemen 1 SBS managemen firms do no necessarily employ he same managers in boh vehicles. 2

3 firm of he invesor). Specifically, muual fund invesors may enrus heir reiremen savings in firms, or even managemen eams, ha hey already know and rus from previous experience. Similarly, pension fund paricipans may coninue invesing he non-earmarked reiremen savings in rusworhy firms. From he supply side poin of view, managemen firms have an opporuniy o grow and gain business due o poenial synergies. Firms may share maerial and human resources (Cici e al., 2010), generae economies of scale and/or scope, and achieve cross-fund learning (Berk and Green, 2004; Brown and Wu, 2016). Addiionally, he SBS managemen in muual and pension funds may cause few conflics due o heir similariies in basic aspecs, such as porfolio managemen, analogous managerial sraegies, and similar invesmen asse universes (Del Guercio and Tkac, 2002). On he oher hand, firms expand SBS arrangemens a manager level as an opporuniy o reain alen (Nohel e al., 2010). Firms may offer he SBS managemen o managers wih ouperformance records o leverage heir successful experience and preven heir deparure o hedge funds (Kosovesky, 2010). Wheher SBS managers are seleced among he bes performers, we expec superior performance in SBS managers. However, wheher managemen firms do no pursue o improve performance and focus on building larger firms o increase fee income, saving personnel coss, he bes managers may no be seleced for he SBS ask. From he managers perspecive, managers may find aracive o engage in he SBS managemen o access a wider and more diversified clienele, and larger funds (Agarwal e al., 2009), increasing heir compensaion. Wih regard o muual fund managers, hey may involve in he managemen of pension funds due o he lower 3

4 redempion risk 2 (long-erm invesmens), he lower volailiy of hese insrumens (Thomas e al., 2014), and he possibiliy of aking less liquid posiions (he invesmen reallocaion in pension funds is scarcer han in oher insrumens -Sialm e al., 2015). On he oher hand, pension fund managers may find aracive he SBS managemen in muual funds because hey may access larger funds and more flexible invesmens, and face lower fee boundaries 3. Neverheless, SBS managers should ake ino accoun he differen naure, purposes (Peraki and Zalewska, 2013), and clienele (Del Guercio and Tkac, 2002) of hese insrumens. While pension fund paricipans aim o form a longerm capial for reiremen and canno recover i unil reiremen 4, muual fund invesors ake shorer-erm invesmen decisions, based on heir needs, he marke changes, and he fund resuls. Consequenly, SBS managers should ake ino accoun hese differenial characerisics and apply specific managemen sraegies in each insrumen; oherwise, he performance of paricipans may be affeced. The sparse financial lieraure on he SBS managemen does no show conclusive resuls abou he influence of his managemen on performance, and is resriced o US muual and hedge funds (Nohel e al., 2010; Cici e al., 2010). Nohel e al. (2010) find ha SBS managers ouperform non-sbs managers in USA muual and hedge funds because his ask is primarily graned o sar performers. Conversely, Cici e al. (2010) sudy his pracice a firm level and find underperformance of SBS muual funds wih regard o heir peers, probably due o he exploiaion of conflics of ineres in muual and hedge funds. 2 Pension fund paricipans may ransfer heir savings o oher pension funds under he fund condiions; however, paricipans canno disinves unil reiremen, excep in he cases included in he legislaion of each counry. 3 Some counries esablish maximum limis on he pension fund fees o ensure he reiremen income of paricipans (Tapia and Yermo, 2008). 4 Some counries allow pension fund paricipans o recover heir savings in excepional circumsances; for example, he deah of he paricipan. 4

5 Given he scarce sudies abou SBS managemen, he moivaion of his work is wofold. Firs, he paper sheds ligh on he SBS managemen effec in he performance of UK equiy muual and pension funds, from boh firm and manager perspecives. Second, we sudy he porfolio holdings buil by SBS managers o handle he muliple fund managemen of wo invesmen vehicles. Specifically, we examine wheher SBS managers build more or less similar porfolios wih regard o non-sbs managers running several funds. Increasing he overlapping porfolio degree faciliaes managemen and may enhance inflows (Blocher, 2016), and performance across funds (Cici e al., 2010); however, wheher SBS managers consider he differen purposes and clienele of muual and pension funds, we would expec lower similariy beween he fund holdings of SBS managers. We carry ou his analysis by applying wo-sep GMM esimaions wih nework insrumens. Our resuls show ha he moivaion of managemen firms o exend heir services o anoher indusry is o achieve larger size raher han o improve performance. A higher firm size ranslaes ino economies of scale and more income (via fees), which allows firms o be more efficien by charging lower managemen fees. On he oher hand, in general, SBS managers do no perform differenly han heir peers, and he performance obained in he pre-sbs period is no significan for he SBS manager performance. Therefore, he main purpose behind SBS arrangemen is no o reain alen, and he SBS managers presen dissimilar skills. Specifically, op (boom) SBS managers under- (ou-) perform heir peers. We furher examine he laes resuls by analysing he invesmen sraegies carried ou o deal wih he muliple fund managemen. We find ha holding more similar porfolios enhances performance; hence, he op non-sbs and op SBS managers develop more similar porfolios among heir funds han he boom managers. Noneheless, he ouperformance of op non-sbs 5

6 manager reveals ha hese managers are beer sock-pickers. On he oher hand, boom SBS managers ouperform heir peers because fund similariies do no influence he performance of boom non-sbs managers. The res of he paper proceeds as follows. In he nex Secion, we presen a marke overview and he daa. Our mehod and empirical resuls are shown in Secion 3. We presen our main conclusions in Secion Marke overview and daa The UK muual fund and pension fund indusries. The absence of SBS managemen lieraure in muual and pension funds, as well as ouside he US marke, lends suppor for our analysis in he Unied Kingdom. The muual and pension fund markes in his counry are characerised by disinc feaures ha make our sudy worhwhile. The UK is he second counry in global asse managemen (following he US). The asses under managemen reached GBP 6.9 rillion in 2016, which represens 373% of he UK GDP (The Invesmen Associaion, 2017). Furhermore, 80% of hese asses are managed by insiuional invesors, and 40% of he insiuional asses are handled by pension funds. The marke concenraion has increased over he las en years, and he op en managemen firms accoun for 55% of he asses under managemen (Europe Economics, 2016). Neverheless, he UK is he only European counry in which he five larges fund firms accoun for less han 40% of he asses, co-exising more han 1,840 managemen firms (The Invesmen Associaion, 2016; EFAMA, 2017a). This environmen enhances compeiiveness among firms, and SBS arrangemens may be an opporuniy o be more efficien and increase he marke share in he wo mos imporan collecive invesmen markes. 6

7 Specifically, he UK pension fund indusry is he second larges pension fund marke wih more han EUR 2.1 rillion under managemen (INVERCO, 2016), which represens 97.4% of he UK GDP (OECD, 2016). The UK pension sysem is he mos disincive of he Anglo-Saxon welfare model, in which public pensions are no large. As a resul, privae pension funds are largely developed and represen a significan porion of he reiremen income. Pension funds are currenly he main invesmen vehicle for household savings (56.2% of household savings were invesed in pension funds and insurance -INVERCO, 2016). The larges asse class in pension funds is equiy (43%), and he domesic equiy pension asses accoun for 40% of he oal equiy exposure (Towers Wason, 2016). On he oher hand, he UK muual fund invesmen (EUR 1.54 rillion in Sepember EFAMA 2017b) is lower han he pension fund invesmen, and he household invesmen in muual funds is only 8.9%. Neverheless, he UK is he fifh European counry in muual fund invesmen wih 3.6% of he worldwide asses. The imporance of he UK muual fund indusry has increased from 1% of he UK GDP in 1964 o 49% of he UK GDP in 2015 (The Invesmen Associaion, 2016). The larges asse class is equiy (54% of he oal invesmen), highlighing he UK equiy funds (24.5% of he oal invesmen) (The Invesmen Associaion, 2016) Daa. Our primary daa is obained from Morningsar Direc and includes he daily reurn, monhly reurn, monhly Toal Ne Asses (TNA), managemen fees, incepion dae, manager hisory, managemen firm name, and quarerly porfolio holdings of he UK equiy muual funds and he UK equiy pension funds from January 1999 o June We exclude index funds (muual and pension funds) for robusness. The sample 7

8 includes 939 muual funds and 366 pension funds. Our samples are free of survivorship bias. Since we sudy he SBS arrangemens from he perspecives of he firms and he managers, we use he fund daa o build firm variables and manager variables. Neverheless, before calculaing hese variables we need he fund performance. We obain he monhly performance of each fund as he monhly four-facor alpha of Carhar (1997). The monhly alphas are obained wih he daily fund reurns of each monh. 5 The risk facors used o obain he four-facor alphas are he daily UK facors developed by Gregory e al. (2013) 6 for he UK marke. We hen classify managemen firms as SBS firms when hey simulaneously marke muual and pension funds. The muual and pension funds analysed are owned by a oal of 329 managemen firms and, beween hese, 88 are SBS firms, 210 firms only marke muual funds, and 31 firms only marke pension funds. We calculae he firm variables as follows. The monhly asses under managemen (size) of each firm are obained as he sum of he monhly asses under managemen of all funds owned by each firm. The remaining monhly firm variables (reurn, performance, managemen fees, fund age, and percenage flows) are he monhly value-weighed average fund variables (reurn, performance, managemen fees, fund age, and percenage flows 7 ) of all funds owned by each firm and monh. On he oher hand, we idenify he SBS managers by combining he muual fund and pension fund daa by unique manager names and look for maches. For each manager name, we obain he manager enure in each fund. If here is an overlap in he 5 The use of daily reurns o obain he monhly alphas ensures a leas 19 observaions by monh, and avoids he drawbacks of esimaing wih rolling windows. 6 All facor daa are obained from he Xfi Cenre for Finance and Invesmen, Universiy of Exeer: hp://business-school.exeer.ac.uk/research/areas/cenres/xfi/research/famafrench/disclaimer/ 7 Percenage fund flows are obained as: Flows i = (TNAi - TNAi-1* (1+ Ri ))/TNAi- 1, where TNA i are he Toal Ne Asses of fund i a monh, and R i is he reurn of fund i a monh. In addiion, flows are winsorized a he boom and op 1% level of he disribuion o ensure ha exreme values do no drive our resuls. 8

9 muual and pension fund managemen periods, we hen classify he manager as a SBS manager during he overlapping period. The sample conains 906 muual-fund managers and 515 pension-fund managers. We clarify ha some funds are side-by-side managed by several managers over ime; as a resul, we mach he fund informaion o he specific manager during her/his specific managemen period. The SBS manager sample is formed by 483 managers, who run 852 muual funds and 325 pension funds for a leas one year in our sample period. We hen calculae he monhly manager variables (reurn, performance, fees, flows, fund age and asses under managemen) for SBS and non-sbs managers by maching he fund informaion wih he managemen period of each manager. Wheher managers (SBS or non-sbs) run several funds, we obain he monhly average reurn, performance, flows, managemen fees, and fund age of all funds managed each monh, and he manager asses are he oal monhly asses under managemen of all funds handled. We should clarify ha we will separaely analyse he SBS manager labour in muual funds, he SBS manager labour in pension funds, as well as he oal SBS managemen. For his reason, we separaely calculae he monhly variables of he SBS managers in muual funds, he monhly SBS manager variables in pension funds, and he aggregae monhly SBS manager variables. Table 1 shows summary saisics of he managemen firms and he managers analysed in panels A and B, respecively. Panel A shows ha SBS firms are more efficien han non-sbs firms (higher reurn and higher performance), and SBS arrangemens allow firms o grow: SBS firms are larger, receive more flows, and heir funds are younger since firms consiue funds when hey ener he oher marke. Furhermore, SBS firms charge lower managemen fees, possible because he larger size allows hem o be more compeiive. Panel B suggess cerain efficiency in he SBS 9

10 arrangemens a manager level. SBS managers presen higher reurn and higher performance, and SBS firms may be saving personnel coss since SBS managers handle older funds; ha is, experienced managers engage in he SBS managemen and coninue managing heir old funds. Addiionally, SBS managers achieve o manage larger porfolios and increase heir compensaion via boh higher managemen fees and larger funds. Comparing non-sbs managers, muual fund managers display higher reurn and higher performance, manage smaller funds, and receive lower managemen fees han pension fund managers. The resuls of SBS managers by marke sugges ha SBS managers are more efficien in he muual fund indusry. Specifically, managers display higher reurn and higher performance han in he pension fund indusry, and handle larger and older muual funds; however, muual fund managemen fees are lower. These indusry differences may help SBS managers o increase heir compensaion hrough wo channels: he larger size of he muual funds managed and he higher managemen fees of pension funds. 10

11 Table 1. Manager summary saisics. Table 1 shows summary saisics of he firms and managers analysed. Panel A shows saisics (reurn, performance -four-facor alpha-, average firm asses in EUR, flows, fund age in monhs, and managemen fees in percenage) of he managemen firms (all firms, non-sbs firms and SBS firms). Panel B shows summary saisics of he managers analysed (reurn, performance -four-facor alpha-, fund asses in EUR, family fund asses in EUR, fund flows, fund age in monhs, and managemen fee in percenage) for all managers, non-sbs managers, SBS managers, non-sbs muual fund managers, non-sbs pension fund managers, and SBS managers separaely in muual and pension funds. Panel A: Firm saisics All Non-SBS firms SBS firms Reurn Performance Average firm asses 2.69* * *10 10 Flows Fund age Managemen fee 1.05% 1.07% 1.01% Panel B: Manager saisics All managers Non-SBS managers SBS managers Non-SBS muual fund managers SBS managers Non-SBS pension fund managers Muual funds Pension funds Reurn Performance Fund asses 1.44* * * * * * *10 8 Family asses 1.48* * * * * * *10 8 Fund flows Fund age Managemen fee 1.08% 1.04% 1.09% 1.04% 1.05% 0.97% 1.24% 11

12 3. Mehodology and resuls Influence of he side-by-side managemen in managemen firms. In his secion we analyse he SBS arrangemens a he managemen firm level. SBS firms may access broader markes and clienele, increase heir size, marke share, and benefis; however, his pracice does no guaranee greaer efficiency and higher performance. We develop model (1) o sudy he influence of SBS arrangemens in he firm performance. i, 0 1SBS i, 2 MFi, 3 i, 1 4 Asses i, 5 Age i, 6 Flows i, 7 Fee i, i, (1) Where: α i, is he 4-facor alpha of managemen firm i in monh. SBS i, is a dummy ha equals one if managemen firm i is a SBS firm in monh, and zero oherwise. We also examine he imporance of specialising in one indusry, and inroduce he MF dummy, which equals one if managemen firm i only markes muual funds and zero oherwise. The following variables are inroduced as conrol variables: α i,-1 is he four-facor alpha of firm i in monh -1; Asses i is he naural logarihm of he asses under managemen of managemen firm i a monh ; Age i is he average age of he funds (in naural logarihm and monhs) of he firm i a monh ; Flows i are he percenage flows of managemen firm i a monh ; Fee i is he average managemen fees charged by he funds of managemen firm i a monh ; and ε i, is he error erm. Table 2 collecs he resuls of model (1), esimaed wih robus sandard errors and clusered by firm. Panel A does no show significanly differen performance beween SBS and non-sbs firms (non-significan SBS coefficiens), and he MF dummy does no show significan differences beween he firms operaing in one or anoher fund indusry. We furher divide boh he SBS and non-sbs firm samples ino performance quiniles o examine performance differences beween firms, since he preliminary resuls of panel A of Table 1 displays higher SBS firm performance. 12

13 Table 2. Performance of SBS managemen firms relaive o heir peers. Table 2 shows he resuls of model (1) esimaed by OLS wih robus sandard errors clusered by firm. All monhly variables are value-weighed average according o he monhly asses under managemen of each firm. Panel A shows he resuls of all firms analysed. Panel B shows he resuls of model (1) by dividing he firm sample ino quiniles according o he firm performance. Quiniles 1 and 5 refer o he boom and op performing firms, respecively. T- saisics are repored in parenheses. *, ** and *** indicae significance a he 10%, 5%, and 1% level, respecively. Panel A: Performance firm resuls (1) (2) SBS (-0.91) (-0.91) MF (0.01) Lagged Alpha *** *** (116.9) (116.9) Asses (0.46) (0.46) Fund age (0.89) (0.89) Managemen *** *** fee (19.98) (19.98) Flows *** *** (-11.86) (-11.86) Consan (-1.09) (-1.09) R Panel B: Firm resuls by performance quiniles Quinile 1 Quinile 2 Quinile 3 Quinile 4 Quinile 5 SBS *** *** (-0.38) (13.78) (0.68) (-19.29) (0.43) MF (0.01) (0.01) (0.01) (0.01) (0.01) Lagged Alpha *** *** *** *** *** (193.69) (-5.32) (-5.07) (11.88) ( ) Asses *** (-0.38) (2.96) (-1.06) (-0.77) (-0.92) Fund age *** *** (0.41) (-2.9) (4.21) (-0.29) (1.12) Managemen *** * *** fee (1.31) (-1.91) (-0.29) (0.19) (529.02) Flows *** *** *** *** ( ) (-1.42) (-8.39) (3.66) ( ) Consan *** *** *** (-0.57) (-13.95) (4.86) (28.79) (-0.81) R

14 Panel B shows ha SBS firms do no perform differenly han heir peers in he firs, hird and fifh quiniles (he firs and fifh quiniles refer o he boom and op performing firms, respecively). However, he second wors (bes) performance SBS firms ouperform (underperform) heir non-sbs firm peers; ha is, SBS agreemens help poor performing firms o a greaer exen. We check wheher hese resuls are driven by size effecs because he 2 nd and 4 h quinile firms are larger han he firms of he remaining quiniles. We divide firms ino size quiniles and conclude ha our resuls are no driven by size effecs since we find he same resuls (ou- and underperformance in he 2 nd and 4 h quinile firms, respecively, and no significan resuls in he remaining quiniles) SBS managers versus non-sbs managers in muual and pension funds. In his secion we analyse he performance of SBS managers wih regard o non- SBS managers. Nohel e al. (2010) argue ha firms develop SBS arrangemens a manager level o reain alen and avoid he deparure of he bes managers o hedge funds (Kosovesky, 2010). Wheher SBS managers are seleced among he bes managers, we expec SBS managers o ouperform non-sbs managers. We sudy separaely he performance of managers in muual and pension funds o examine wheher SBS managers behave differenly in each indusry. Addiionally, hese wo disinc analyses allow us o deermine he exisence of SBS favouriism (Nohel e al., 2010; Cici e al. 2010); ha is, wheher SBS managers favour one of he invesmen vehicles a he expense of he oher. The mos common favouriism source is he dissimilar remuneraion beween insrumens. However, we do no expec his ype of conflics because he remuneraion of muual- and pension-fund managers depends on managemen fees, which are obained as a percenage of asses under managemen in 8 These resuls are no shown for breviy and are available upon reques. 14

15 boh muual and pension funds. Despie of his, we noice ha several counries impose pension fund fee limis o ensure reiremen income (Tapia and Yermo, 2008), which may provoke SBS managers o maximize fee income by ransferring resources from pension funds o muual funds. In he case of he UK, only some ypes of pension funds presen specific regulaion; for insance, he occupaional pension funds of he auomaic enrolmen pension ensure a cap of 0.75% per year of he AUM (Policy Saemen 15/5 Financial Conduc Auhoriy, he Occupaional Pension Schemes Regulaions 2015, regulaions 6 and 7). Wheher SBS managers favour muual funds a he expense of pension funds o overcome pension fee consrains, SBS managers will ouperform non-sbs muual fund managers and underperform non-sbs pension fund managers. We include he prior consideraions in model (2), following a similar mehodology o ha applied by Nohel e al. (2010). 6 i, 0 1SBS i, 2MForigini, 3 Pr e SBSi, 4Pos SBSi, iconroli, i, i 1 (2) Where α i, is he four-facor alpha of manager i in monh. SBS i, equals one if manager i is a SBS manager in monh, and zero oherwise. MForigin i, is a dummy variable ha conrols he origin indusry of he manager, and equals one wheer SBS manager i sared in he muual fund indusry, and zero oherwise. A pre-sbs period dummy is also included o analyse he influence of he previous experience in he SBS managemen, and o examine wheher SBS arrangemens pursue o reain alen. Pre- SBS i, equals one if manager i is no a SBS manager in monh and he/she becomes a SBS manager some poin afer, and zero oherwise. We deec ha some managers leave he SBS managemen and come back o manage funds in one of he indusries analyzed. Wheher hese managers learn from he crossed-indusry managemen, hey will presen superior performance afer he SBS 15

16 period. We hen inroduce he Pos-SBS i, dummy, which equals one if manager i is no a SBS manager in monh and was previously a SBS manager, and zero oherwise. The conrol variables of he model are: α i,-1 (he four-facor alpha of manager i in monh - 1), Asses i, (he naural logarihm of he asses under managemen of manager i a monh ), Famasses i, (he naural logarihm of he family firm asses of manager i a monh ), Age i, (he naural logarihm of he average age in monhs of he funds handled by manager i a monh ); Flows i, (he percenage money flows of manager i a monh ); Fee i, (he average managemen fee received by manager i a monh ). Finally, ε i, is he error erm. We compare he model (2) resuls of SBS managers in muual funds wih regard o non-sbs muual fund managers in Table 3. Panel A shows ha SBS managers in muual funds do no perform differenly han non-sbs muual fund managers (nonsignifican SBS dummy). Therefore, we do no find evidence ha sar managers engage in he SBS managemen. Addiionally, he background of SBS managers does no influence manager performance (non-significan MForigin and pre-sbs). Specifically, he SBS managers ha sared in he muual fund indusry do no perform differenly han hose who sared in pension funds. Nohel e al. (2010) argue ha he manager s origin is no imporan if he moivaion of he SBS managemen is o exploi poenial conflics of ineres in he SBS relaionship. Consequenly, our resul does no allow us o discard some favouriism from pension funds o muual funds, as Cici e al. (2010). On he oher hand, he prior non-sbs performance does no influence he SBS manager performance (non-significan pre-sbs dummy); hence, SBS arrangemens do no pursue o reain alen. Consisen wih his resul, Cici e al. (2010) find no evidence ha SBS arrangemens arac and reain superior alen, and SBS managers are no necessarily he bes performers. This may be due o several reasons. Firs, managemen 16

17 firms may offer he SBS posiions o skilful managers in he firs place, and wheher some of hem rejec i, firms offer his managemen o oher managers (wih poorer skills). Furhermore, wheher he purpose of he SBS managemen is o develop favouriism pracices, no skilled managers are needed. As a resul, SBS managers will presen differen skill levels. We also find ha managers who leave he SBS managemen and subsequenly run muual funds ouperform (significanly posiive pos-sbs). This evidences some learning from he SBS experience. Conversely, Cici e al. (2010) find ha US muual funds do no benefi from he SBS affiliaion wih hedge funds. These apparen conrary resuls sugges ha SBS arrangemens in muual and pension funds are more effecive han in oher indusries, probably because of he similariies beween hem. We furher analyse wheher SBS managers presen diverse skills. We divide he SBS and non-sbs manager samples ino performance quiniles. Panel B of Table 3 shows he resuls of model (2) by performance quiniles (he resuls of he conrol variables are no displayed and are available upon reques). We find ha op SBS managers (5 h quinile) do no perform differenly han heir peers, and SBS managers ouperform heir peers in he remaining quiniles. Tha is, managers wih diverse skills (usually no he bes managers) engage in his pracice. 17

18 Table 3. SBS managers in muual funds relaive o non-sbs muual fund managers. Table 3 is divided ino wo panels. Panel A shows he resuls of model (2) for all muual fund managers analysed. Panel B shows he model (2) resuls by dividing he SBS and non-sbs muual fund manager samples ino quiniles according o he monhly manager performance of SBS and non-sbs managers, respecively. The resuls of conrol variables are no displayed in Panel B and are available upon reques. Quiniles 1 and 5 refer o he boom and op performers, respecively. All models are esimaed wih syle fixed effecs, robus sandard errors, and clusered by manager. T-saisics are repored in parenheses. *, ** and *** indicae significance a he 10%, 5%, and 1% level, respecively. Panel A: Performance resuls (1) (2) (3) (4) (5) SBS (0.25) (0.11) (0.25) (-0.44) (-0.58) MForigin (0.43) (-0.43) Pre-SBS (0.01) (0.01) Pos-SBS *** ** (2.9) (2.31) Lagged Alpha *** *** *** *** *** (14.12) (14.12) (14.12) (14.1) (14.08) Asses * * * (-1.68) (-1.63) (-1.68) (-1.73) (-1.56) Fam_asses *** *** *** *** ** (2.9) (2.89) (2.9) (2.95) (2.55) Fund age (-0.32) (-0.32) (-0.32) (-0.23) (-0.21) Flows * * * * * (-1.71) (-1.73) (-1.71) (-1.81) (-1.77) Mmgm fee (1.17) (1.17) (1.17) (1.2) (1.16) Consan (0.34) (0.32) (0.34) (0.38) (0.43) R Syle effec Yes Yes Yes Yes Yes Panel B: Resuls by manager performance quiniles Quinile 1 Quinile 2 Quinile 3 Quinile 4 Quinile 5 SBS ** *** *** *** (2.27) (6.82) (3.69) (-4.94) (-1.59) MForigin * (0.13) (-0.31) (0.04) (-1.65) (-0.16) Pre-SBS (0.005) (0.001) (0.001) (0.001) (0.001) Pos-SBS * (1.63) (-1.84) (0.78) (0.59) (-1.38) Conrol var. Yes Yes Yes Yes Yes R Syle effec Yes Yes Yes Yes Yes 18

19 We repea he prior analysis for pension funds in Table 4. Panel A does no display a differen performance beween SBS and non-sbs managers (SBS dummy). The origin of SBS managers is imporan in he pension fund indusry, and managers saring in muual funds underperform. This reveals ha muual fund managers find some difficulies in adaping o he he pension fund indusry. Furhermore, hese resuls do no confirm he exisence of favouriism from pension o muual funds (suggesed in Table 3). In fac, he exisence of SBS favouriism in he financial lieraure is unclear; for insance, Nohel e al. (2010) do no find evidence of favouriism in SBS managers. On he oher hand, he pre-sbs dummy is significanly negaive (only in column 3), revealing an inverse relaion beween he pre-sbs performance and he SBS performance. Tha is, he SBS managemen helps poor performers and damages good performers. This odd resul indicaes ha managers wih dissimilar skills paricipae in he SBS managemen, and why op performers may no be ineresed in SBS pracices (he SBS managemen requires exra effor and aenion, and good performers diver from heir focus, eroding he performance). Finally, SBS managers who leave he SBS managemen and subsequenly manage pension funds do no perform differenly (nonsignifican pos-sbs). This resul conrass wih he evidence found in muual funds (Table 3), hus, SBS managers who become non-sbs managers only saisfacory apply heir SBS knowledge in he muual fund indusry. We also divide he SBS and non-sbs manager samples ino performance quiniles o furher analyse dissimilar managerial skills. Panel B does no show differences beween SBS and non-sbs managers in any quinile (excep for a significanly posiive pos-sbs coefficien in he firs quinile). The loss of significance wih regard o Panel A resuls indicaes more homogeneous SBS and non-sbs manager samples in pension funds. 19

20 Table 4. SBS managers in pension funds relaive o non-sbs pension fund managers. Table 4 is divided ino wo panels. Panel A shows he resuls of model (2) for all pension fund managers analysed. Panel B shows he model (2) resuls by dividing he SBS and non-sbs pension fund manager samples ino quiniles according o he monhly manager performance of SBS and non-sbs managers, respecively. The resuls of conrol variables are no displayed in Panel B and are available upon reques. Quiniles 1 and 5 refer o he boom and op performers, respecively. All models are esimaed wih syle fixed effecs, robus sandard errors, and clusered by manager. T-saisics are repored in parenheses. *, ** and *** indicae significance a he 10%, 5%, and 1% level, respecively. Panel A: Performance resuls (1) (2) (3) (4) (5) SBS (0.97) (0.97) (0.97) (0.97) (0.97) MForigin ** ** (-2.28) (-2.4) Pre-SBS ** (-2.28) (0.17) Pos-SBS (1.05) (1.07) Lagged Alpha *** 0.084*** 0.084*** *** *** (17.23) (17.22) (17.22) (17.24) (17.23) Asses * * * * * (-1.67) (-1.91) (-1.91) (-1.7) (-1.95) Fam_asses *** *** *** *** *** (4.28) (4.5) (4.5) (4.31) (4.56) Fund age (1.59) (1.55) (1.55) (1.49) (1.43) Flows (0.56) (0.97) (0.97) (0.41) (0.83) Mmgm fee (-0.27) (-0.15) (-0.15) (-0.34) (-0.21) Consan * ** ** * ** (-1.96) (-2.05) (-2.05) (-1.94) (-2.04) R Syle effec Yes Yes Yes Yes Yes Panel B: Resuls by manager performance quiniles Quinile 1 Quinile 2 Quinile 3 Quinile 4 Quinile 5 SBS (0.07) (0.07) (0.07) (0.07) (0.07) MForigin (-0.34) (-1.56) (-1.24) (-0.82) (-0.03) Pre-SBS (0.11) (0.11) (0.11) (0.11) (0.11) Pos-SBS *** (3.2) (-0.48) (0.06) (-0.47) (-0.07) Conrol var. Yes Yes Yes Yes Yes R Syle effec Yes Yes Yes Yes Yes 20

21 3.3. SBS managers versus non-sbs managers. Robusness ess. In his secion we examine he overall SBS manager performance wih regard o he non-sbs manager performance, wihou differeniaing beween muual and pension fund oucomes. We merge he SBS manager variables in muual funds and pension funds o obain he aggregae SBS manager variables 9. Table 5 presens he resuls of model (2). Panel A shows ha SBS managers do no perform differenly han non-sbs managers, he background in muual funds and he pre-sbs experience do no influence performance, and he SBS period has posiive repercussions on he pos-sbs performance. Wheher managers learn from he SBS experience, firms migh use he SBS managemen o rain managers and enhance managerial skills ha would be used o ouperform in he pos-sbs managemen. In view of he resuls from Tables 3-5, he reaining alen hypohesis loses srengh; o be precise, firms do no only offer SBS arrangemens o he bes managers, and/or he bes managers are no ineresed in hese agreemens if he exra effor is no sufficienly rewarded and heir performance is affeced. On he oher hand, wheher managers implemen successfully he knowledge acquired during he SBS period, managers may engage in his experience o promoe heir careers and employ he SBS managemen as evidence of addiional skills and apiudes. 9 To obain he resuls of each SBS manager (performance, fund size, fund age, flows and fees), we follow he process explained in secion 2. 21

22 Table 5. Performance of SBS managers relaive o heir peers. Table 5 is divided ino wo panels. Panel A shows he resuls of model (2) for all managers analysed. Panel B shows he model (2) resuls by dividing he SBS and non-sbs managers ino quiniles according o he monhly performance of SBS and non-sbs managers, respecively. The resuls of conrol variables are no displayed in his panel and are available upon reques. Quiniles 1 and 5 refer o he boom and op performers, respecively. All models are esimaed wih syle fixed effecs, robus sandard errors, and clusered by manager. T-saisics are repored in parenheses. *, ** and *** indicae significance a he 10%, 5%, and 1% level, respecively. Panel A: Performance resuls (1) (2) (3) (4) (5) SBS (0.63) (0.3) (0.63) (0.08) (0.6) MForigin (1.41) (0.89) Pre-SBS (0.17) (0.17) Pos-SBS ** ** (2.47) (2.29) Lagged Alpha *** *** *** *** *** (12.95) (12.94) (12.95) (12.95) (12.94) Asses *** *** *** *** *** (-3.45) (-3.35) (-3.45) (-3.53) (-3.45) Fam_asses *** *** *** *** *** (5.71) (5.65) (5.71) (5.75) (5.71) Fund age (0.48) (0.59) (0.48) (0.63) (0.68) Flows (-1) (-1.04) (-1) (-1.1) (-1.11) Mmgm fee (-0.58) (-0.59) (-0.58) (-0.59) (-0.6) Consan ** ** ** * ** (-1.98) (-1.97) (-1.98) (-1.92) (-2.03) R Syle effec Yes Yes Yes Yes Yes Panel B: Resuls by manager performance quiniles Quinile 1 Quinile 2 Quinile 3 Quinile 4 Quinile 5 SBS *** *** *** *** *** (8.48) (29.29) (16) (-10.72) (-5.96) MForigin (-0.15) (0.39) (-0.29) (-1.6) (0.62) Pre-SBS (0.01) (0.01) (0.01) (0.01) (0.01) Pos-SBS ** (2.23) (-1.33) (1.49) (-0.08) (-0.53) Conrol var. Yes Yes Yes Yes Yes R Syle effec Yes Yes Yes Yes Yes 22

23 The analysis by performance quiniles (panel B) confirms ha op managers usually remain in he non-sbs managemen. The underperformance of op SBS managers (4 h and 5 h quiniles) displays limied skills o handle muliple funds of differen naure wih regard o op non-sbs managers. Noneheless, middle and boom SBS managers (1 s, 2 nd and 3 rd quiniles) ouperform heir non-sbs peers, which poins ou ha modes managers venure ino SBS managemen because of he poenial benefis (managemen of larger funds and addiional managemen fee reward). As an addiional analysis, we develop model (3) o sudy wheher SBS managers presen specific-indusry skills, and achieve beer oucomes in one indusry. In his model we only examine SBS managers, and compare he performance obained in muual funds wih he performance obained in pension funds. 6 i, 0 1SBS _ MFi, 2MForigin i, iconrol i, i, i 1 (3) Where: SBS_MF equals one for he resuls obained by SBS managers in muual funds, and zero oherwise. MForigin equals one if he manager sared in he muual fund indusry, and zero oherwise. The conrol variables are defined in (2). Table 6 does no display significan evidence ha SBS managers are beer in one of he indusries analysed (non-significan SBS_MF). The manager background is no significan eiher. Accordingly, SBS managers are able o handle funds of disincive naure, and he main ineres behind hese arrangemens is o manage larger porfolios and increase heir fee remuneraion (significanly posiive coefficiens in asses, family asses, and managemen fee variables). 23

24 Table 6. Performance of SBS managers in muual and pension funds. Table 6 shows he resuls of model (3). All models are esimaed wih syle fixed effecs, robus sandard errors, and clusered by manager. T-saisics are repored in parenheses. *, ** and *** indicae significance a he 10%, 5%, and 1% level, respecively. (1) (2) SBS_MF dummy (-0.62) (-0.52) MForigin dummy (-0.59) Lagged Alpha *** *** (19.21) (19.21) Asses ** ** (-2.22) (-2.27) Fam_Asses *** *** (4.43) (4.46) Fund age (0.58) (0.58) Flows (-0.11) (-0.06) Managemen fee *** *** (6.85) (6.85) Consan (-0.87) (-0.85) R Syle effec Yes Yes 3.4. Sraegies for and effec of managing muliple funds: he fund similariies beween he funds of one manager. In his secion we furher examine he reasons behind he opposie performance resuls of boom and op SBS managers (ou- and under-performance, respecively), found in Secion 3.2. We analyse wheher boom and op SBS managers develop dissimilar managerial sraegies o handle he muliple fund managemen of wo differen naure invesmen vehicles, and how hese sraegies influence flows and performance. We previously argue ha SBS managers are no usually he bes performers, and he limied skills o handle several funds of differen naure may explain he 24

25 underperformance of op SBS managers. The muliple fund managemen requires exra effor, bu he ime and aenion of managers are limied. As a resul, managers may faciliae heir managemen by increasing he overlapping porfolio degree of heir funds (Cici e al., 2010). We hen hypohesize ha op SBS managers consider he differen naure of muual and pension funds and build porfolios wih higher asse variey han non-sbs managers, which leads hem o underperform. On he conrary, we expec ha boom SBS managers ouperform because hey presen higher overlapping porfolio degree han boom non-sbs managers. In his analysis, we resric our sample o he op (fifh quinile) and boom (firs quinile) SBS managers and he op and boom non-sbs managers (fifh and firs quiniles, respecively). In addiion, we require non-sbs managers o handle several funds o es our argumen abou he porfolio similariies in muliple fund managemen. Furhermore, our daa source only provides fund holdings from March 2002; herefore, he ime period of his analysis is from March 2002 o June The resuling op manager sample is resriced o 93 SBS managers and 25 non- SBS managers (all non-sbs managers are muual fund managers), and he boom manager sample is formed by 94 SBS managers and 74 non-sbs managers (all of hem muual fund managers). The non-sbs manager samples are more limied because we exclude managers handling one fund 10. The op managers handle 201 muual funds and 104 pension funds. The boom managers run 162 and 97 muual and pension funds, respecively. We clarify ha some funds are consecuively handled by differen managers. In hese cases, he fund daa is assigned o his/her manager during his/her managemen period. The fund disribuion among managers shows ha he op SBS managers run 152 and 104 muual and pension 10 The op non-sbs manager quinile was formed by 83 managers and he boom non-sbs manager quinile was formed by 84 managers; herefore, many op non-sbs managers focus on running one fund. 25

26 funds, respecively; and he non-sbs managers run 49 muual funds. Boom SBS managers run 149 muual funds and 94 pension funds, and boom non-sbs managers run 142 muual funds (no pension funds). Once we selec he manager sample and idenify he funds handled by each manager in each monh, we examine he porfolio holdings of he seleced funds 11 o deermine he similariy holding levels beween all he funds run by each manager. Afer ha, we sudy he influence of he fund similariies in he flows and he performance of all he funds handled by one manager. We develop a similariy fund measure based on he similariy measure developed by Blocher (2016). This auhor creaes a measure ha compares he porfolio holdings of wo conneced muual funds. Deparing from Blocher s measure, we compare he holdings of all funds handled by one manager in pairs, and obain he similariy measures for all possible fund pairs. In a second sep, we also include he similariies of he remaining funds handled by one manager in he preliminary measure. This process is repeaed for all managers sudied. The similariy measure is buil as follows. We build porfolio vecors (h i, ) wih he weighs of each securiy held by fund i in quarer. If here are m securiies, each vecor is m x 1 every quarer. The similariy measure beween funds i and j managed by one manager a quarer (s ij, ) is obained from he porfolio vecors h i, and h j,. The similariy measure of funds i and j (s ij, ) is he do produc beween h i, and h i, divided by he Euclidean norm produc of each vecor: s ij, hi, h j, (4) h h i, j, Where he Euclidean norm h i, is defined across m securiies as: 11 The funds analysed have a leas 75% of heir holdings invesed in equiy. 26

27 M 2 h i, him, (5) m 1 The similariy measure ranges from 0 o 1, indicaing no porfolio overlapping and idenical porfolios, respecively. The similariy measure is calculaed for all possible fund pairs among all he funds run by each manager analysed every quarer. We hen obain a peer-weighed similariy measure beween funds i and j considering he similariies of he k funds managed by he analysed manager a quarer : sij, PeerWeigh ij,, k i, j i (6) s k ik, Subsequenly, we capure he influence of he PeerWeigh similariy in he fund flows and he fund performance wih ineracive peer measures (PeerFlow and PeerAlpha). Neverheless, firs, we should clarify ha all our fund daa is monhly, excep for he porfolio holdings, which are quarerly 12. We hen relae he monhly flows and he monhly performance wih approximaed monhly peer measures, following he approximaion mehod applied by Brown e al. (2018). Specifically, he monhly flows and he monhly performance are mached wih he mos recenly quarerly PeerWeigh compued (e.g., a PeerWeigh obained a he end of March is used o calculae he peer measures in April, May and June, -Brown e al. 2018). Therefore, he PeerMeasure i, of fund i a monh considering he k funds managed by one manager is defined as follows: PeerMeasur ei, PeerWeigh Measure where k i k ik, k,, : (7) where: PeerMeasure i is he PeerFlow i or he PeerAlpha i of fund i a monh (depending on wheher he flows or he performance is examined); Measure k, is he 12 Regulaion only requires quarerly holding repor. Addiionally, alhough a few funds repor higher frequency holdings in our daabase, hese are uneven and a differen ime basis. 27

28 Flow k, (percenage money fund flows 13 ) or he Alpha k, (four-facor alpha) of he remaining k funds (excluded fund i) handled by he manager analyzed a monh. The peer measures accumulae he peer weighed average similariy of flows/performance ino or ou of he funds run by one manager. We hen apply a wo-sep generalised mehod of momens (GMM) wih nework insrumens (Kelejian and Prucha 1998; Blocher 2016) o examine he effec of he porfolio holding similariies in he flows and he performance. This mehod overcomes he endogeniy of he peer measures (he average flows/alpha of one fund affecs he oher managers funds, and he flows/alpha of he oher funds affecs ha fund). Addionally, his mehod idenifies he rue exernaliy effec beween funds because i disinguishes fund exernaliies from common exogenous fund characerisics and correlaed effecs, such as fund age or fund size (Blocher, 2016). The firs GMM sep is also divided ino wo sages and idenifies he insrumens ha predic he endogenous variables (flows and alpha). The model of he firs sage is: 2, i Insrumen s 0 i 1 PeerMeasur e (8) i Where: PeerMeasure i is he PeerFlow i or he PeerAlpha i of fund i a monh, and he Insrumens are he lagged peer-flow or he lagged peer-alpha, compued wih he conemporaneous peer-weighs a and he lagged flow/alpha a -1; ha is: Lagged _ i k ik k 1 PeerMeasur e PeerWeigh Measure,,,, k i, and he square of he 2 lagged peer flow/alpha ( Lagged _ PeerMeasur e i ) o provide over-idenificaion., We obain he prediced values of he PeerFlow ( PeerFlow ) and he PeerAlpha ( PeerAlpha ) from model (8) and include hem ino he model of he second sage: 13 The fund flows are: Flows i = (TNA i - TNA i -1 * (1 + R i ))/TNA, where TNA i -1 i are he Toal Ne Asses of fund i a monh and R i is he reurn of fund i a monh. To ensure ha exreme values do no drive our resuls, flows are winsorized by fund a he boom and op 1% level of he disribuion. 28

29 Measure i, PeerMeasur ei, iconrolvar iables (9) 5 i 1 Where: Measure i, is he Flow i, or he Alpha i, of fund i a monh ; PeerMeasur, is he e i PeerFlow or he PeerAlpha ; and, following Blocher (2016), he conrol variables are: he lagged fund flows, he lagged fund alpha, he lagged fund size, he lagged fund age, and he lagged fund volailiy of fund i. The lagged variables are included in ha form o avoid spurious correlaion effecs. Clarify ha he fund alpha is he monhly four-facor fund alpha obained from he daily fund reurns each monh; he fund size is he logarihm of he oal ne asses of fund i; he fund age is he logarihm of he fund age in monhs, and he fund volailiy is he monhly volailiy of he fund reurns obained from daily fund reurns each monh. Following he prior procedure, we calculae he prediced flows ( Flow) and he prediced alpha ( Alpha ) from model (9), and include hem in he PeerMeasure (7); ha is, in he PeerFlow and PeerAlpha, respecively, obaining he prediced (Flow) Peer and Peer ( Alpha ). Afer ha, he prediced insrumens in he second sep. (Flow) Peer and Peer ( Alpha ) are used as The second sep is also divided ino wo sages. The models of he firs sage are: PeerFlow i, = β0peer( Flow) i, + ε0 (10) i, 0 Peer Alpha) i, PeerAlpha ( (11) PeerFlow Similarly, we compue he prediced values of he PeerFlow ( 0 i, ) and PeerAlpha he PeerAlpha ( i, ) from models (10) and (11), respecively. Finally, we esimae he models of he second sage: 5 i, PeerFlow i, iconrolvar iables i i 1 Flow, (12) 29