Centre on Regulation and Competition WORKING PAPER SERIES

Transcription

1 Cente on Regulaton and Competton WORKING PAPER SERIES Pape No. THE EFFICIENCY OF PENSION FUND MANAGERS IN LATIN AMERICA Amando Baentos Unvesty of Mancheste Azz Boussofane Unvesty of Hetfodshe Busness School Octobe 2 ISBN: Futhe detals: Publshed by: Fona Wlson, Cente Secetay Cente on Regulaton and Competton, Insttute fo Development Polcy and Management, Unvesty of Mancheste, Cawfod House, Pecnct Cente, Oxfod Road, MANCHESTER M3 9GH Tel: Fax: Emal: cc@man.ac.uk Web:

2 THE EFFICIENCY OF PENSION FUND MANAGERS IN LATIN AMERICA INTRODUCTION The spead of penson efom n Latn Ameca n the 99s has eplaced unfunded sem-publc penson schemes wth ndvdual etement savng plans managed by pvate penson fund manages (Baentos 998; Mesa-Lago and Betanou 998; Quesse 998b). The ceaton of a penson fund management maket s one of the most consequental effects of the efom, wth fa eachng mplcatons fo the evoluton of the efom, and ts mpact upon the economes of the egon (Baentos 999). Penson fund manages have many mpotant functons. They collect wokes' contbutons nto a penson fund, nvest ths fund n a ange of fnancal assets, aange dsablty and suvvo nsuance fo actve contbutos, penson benefts fo those eted, and povde a ange of suppotng sevces. The success o falue of the penson efom wll come to depend on whethe the penson fund manages ae effectve as penson povdes. The desgnes of the new penson schemes n Latn Ameca amed fo the penson fund manages to opeate n a compettve envonment, albet subject to detaled supevson and egulaton. Competton extends to commsson and chages, ates of etun, and the qualty of the sevces povded. Contbutos ae able to tansfe easly fom one penson fund manage to anothe n seach of a bette deal. Ths s meant to ensue that compettve pessues spu the penson fund manages. The egulaton of fund manages s set by govenment authotes, and mplemented by an ad hoc egulato. Regulaton s extensve, and coves poducts, commsson and chages, nvestment potfolos, ates of etun, and standads of sevce and pobty. By compason wth ssues of penson plan desgn, the lteatue on penson fund manages n Latn Ameca s small (Quesse 997, 998a; Snha, Matnez et al. 998; Baentos 999). In pat ths s because penson efom has been establshed only ecently n the egon. Gven the pefomance of penson fund manages to date has been mxed, ssues egadng the effcency Penson efom was mplemented n Chle (98), Peu (993), Agentna and Colomba (994), Uuguay (996), Bolva and Mexco (997), Costa Rca and El Salvado (998). Smla efoms have been mplemented n Poland, Hungay and Kazakhstan. 2

3 and egulaton of the penson fund management maket must be tackled ugently. The pape ams to contbute to ths lteatue by developng measues of the effcency of penson fund manages usng data envelopment analyss. It dentfes tends n the pefomance a nd effcency of penson fund manages, the mpact of the egulaton, and dscusses whethe maket pessues o egulaton explan the hgh costs of penson plans. The empcal wok focuses on Chle because t has the longest expeence wth penson efom, and because t povded a model of the compettve and egulatoy stuctue fo penson fund manages whch othes countes have followed closely. The pape s ogansed as follows. Secton One povdes an outlne of the man featues of the new penson fund management maket establshed by penson efom. Secton Two explans the focus on the effcency of the new penson fund manages and dentfes the key ssues fo analyss. Secton Thee outlnes data envelopment analyss and dscusses ts appopateness to the study of penson fund manages. Secton Fou pesents estmated measues of effcency of penson fund manages n Chle and dscusses the mplcatons.. Penson Refom and the New Penson Fund Manages The 98 penson efom n Chlean eplaced unfunded penson schemes wth ndvdual etement savng plans. In the new penson plans, dependent wokes ae equed to contbute a facton of the eanngs to an ndvdual etement account wth a penson fund manage. Independent wokes can make voluntay contbutons. The penson fund manages (Admnstadoas de Fondos de Pensones o AFPs) supply the admnstaton, nvestment and management of ndvdual etement accounts. The penson fund manages ae copoatons set up wth the exclusve objectve of managng ndvdual etement accounts, and ae egulated by the Supentendenca de Admnstadoas de Fondos de Pensones ( SAFP). The penson fund manages collect wokes' contbutons nto a penson fund and nvest ths fund n a ange of pemtted assets. Untl ecently, the penson fund manages could only admnste a sngle penson fund fo all the afflates, but egulatons ntoduced n 999 allowed them to offe a second fund wth a fxed nteest potfolo amed at potectng wokes close to etement fom the fluctuatons of vaable nteest assets. Both these funds ae nsulated 3

4 fom the penson fund manages' own assets, whch belong to the shaeholdes to whom pofts ae dstbuted. The penson funds, on the othe hand, ae owned by the afflates. The only flows allowed between the penson fund and the penson fund manages' assets consst of pemtted deductons fo commsson. 2. The Pefomance of Penson Fund Manages: Effcency and Regulaton Issues Ths secton examnes the behavou and pefomance of fund manages aganst n the context of competton and egulaton, and dentfes the man effcency ssues to be nvestgated n the next secton. The effcency of the penson fund manages s cucal to the success of penson efom n Latn Ameca, and wll detemne ts mpact upon the economes of the egon. The efoms have had to establsh a maket fom scatch, as only Bazl had a sgnfcant penson fund management maket befoe the ecent efoms (Yemo 2). Ths, and the faglty of othe fnancal ntemedaes, explans the need fo extensve egulaton of the emegng maket. The pessues of competton and egulaton lagely explan the maket behavou and pefomance of the penson fund manages. The penson fund manages ae expected to compete fo busness, by attactng and etanng saves. At the same tme they ae subject to extensve egulaton, estctng the sevces to standad poducts, as well as detemnng the ncdence, though not the level, of commssons. The egulatoy famewok also ncludes pudental egulaton on penson funds nvestment potfolos, mnmum ates of etun, and standads fo nfomaton dsclosue to both contbutos and the egulato. Competton n the penson fund management maket Standad maket ndcatos povde conflctve sgnals on the extent to whch the penson fund management maket s compettve. Table shows key pefomance ndcatos. 4

5 Table. Key Indcatos of Penson Fund Management Maket n Chle Indcato Numbe of AFPs Lagest 3 shae (contbutos) Afflates (m) Contbutos (m) Tansfes (m) Sales pesonnel (,s) Annual ate of etun Data Souce: Baentos (998), Supentendenca de Admnstadoas de Fondos de Pensones (seveal ssues). Fstly, vaaton n the numbe of penson fund manages ove tme suggests baes to enty eased sgnfcantly n the 99s. Enty equements fo penson fund manages nclude a mnmum captal, a eseve of pecent of the penson fund 2, and 'ft and pope' lcensng by the SAFP (Yemo 2). The numbe of povdes emaned stable thoughout the 98s, oscllatng between 2 and 4. In 992 and 993 the numbe of penson fund manages n the maket ose to 22, but subsequent meges educed these to 3 by August 997. A futhe consoldaton of the maket led to the numbe of penson fund manages fallng to 8 by Decembe 999. The se n the numbe of povdes n the ealy 99s was encouaged by hgh poft levels, and by the elaxaton of enty egulatons. The evoluton n the numbe of penson fund manages suggests that maket enty s not too dffcult, and that aggessve maketng can ntally secue maket shae. At the same tme, the apd declne n the numbe of povdes fom 22 n 994 to 3 n 997 and to 8 n 999 llustates the dffcultes faced by new entants n potectng the maket shae. The consoldaton of the maket n the late 99s esulted fom stong maket pessues and advese nvestment etuns. Secondly, the numbe of penson plan contbutos tansfeng the accounts fom one penson funs mange to anothe suggests a healthy measue of competton. The desgn of ndvdual etement penson plans n Chle encouages ndvdual choce of penson fund manage, as a 2 Ths eseve, o encaje, was ntally 5 pecent of the penson fund. 5

6 means of ensung that compettve foces peval. In ode to facltate ths choce, penson poducts ae standad, commssons ae unfom fo all afflates to a sngle AFP, and seconday poducts ae closely egulated. In theoy, penson plan patcpants n Chle ae expected to compae ates of etun, commsson levels and sevce qualty acoss the dffeent penson fund manages, and tansfe the accounts to the one wth the best deal. Intally, patcpants wshng to tansfe to anothe AFP had to make a equest n peson at a banch offce, but n Febuay 988 the egulatons wee elaxed to allow patcpants to equest a tansfe though a sgned fom. Ths esulted n a apd se n tansfes, futhe encouaged n 992 by the aggessve maketng effots of new maket entants. 3 In 987, thee wee 8.48 tansfes, o pecent of actve contbutos, by 995 thee wee.32 mllon tansfes, o 53 pecent of actve contbutos. Renstatement of the equement to apply fo the tansfe n peson at a banch offce educed tansfes to.48 mllon n 999, o aound 8 pecent of actve contbutos. It s questonable whethe the hgh level of tansfes s a good ndcato of healthy competton. Abuhadba (994) studed the detemnants of tansfes usng a panel of monthly coss sectons of ndvdual tansfes acoss penson fund manages fo the peod Apl 992 to June 993. He egessed tansfes on a ange of ates of etun, sales and commsson vaables. He found that ates of etun and commssons had a small mpact on tansfe behavou, although etuns ove a longe peod had lttle effect, but that the domnant facto was the numbe of sales pesonnel deployed by fund manages. Penson plan patcpants appea to be faly nsenstve to dffeences n pefomance ndcatos acoss penson fund manages. Ths may be due to the hgh nfomaton costs equed to make penson plan choce decsons, 4 o to that fact that these pefomance dffeentals acoss penson funds ae too small to matte. 5 Thdly, maket concentaton s hgh, suggestng that competton s lmted. The thee lagest penson fund manages accounted fo 63 pecent of actve contbutos n 982, sng to 67 pecent n 99, and to 75 pecent n 999, afte a peod of maket consoldaton. Penson fund 3 Tansfe chages ae pemtted by the egulatons but have seldom been appled by penson fund manages. 4 See Valdés-Peto (992) fo a dscusson of the complexty of the nfomaton equed, and Damond (993) fo a moe geneal dscusson of maket falue n the context of nsuance makets. 6

7 management maket concentaton n Chle can be explaned by economes of scale n the management of etement accounts. 6 The geneal lteatue suggests that economes of scale ae pesent n bankng and othe fnancal nsttutons. A ecent suvey of eseach n ths aea concluded that the aveage cost cuve of bankng nsttutons shows a flat U-shape (Bege, Hunte et al. 993). It epots that among US banks wth ove US $ bllon n assets, aveage costs appea to be lowest n the ange of US $2- bllon. In Decembe 995, penson fund manages penson fund assets anged fom US $ 52 to $5 mllon. Mtchell (996) studes the cost stuctue of dffeent penson savng nsttutons and fnds sgnfcant economes of scale n penson fund povson. Fo employe sponsoed penson plans n the USA, costs se by.27 pecent fo evey pecent se n assets, and.8 pecent fo evey pecent se n contbutos. 7 Manda (994) measued economes of scale n the Chlean penson fund management maket. Usng data fom a panel of penson fund manages fo the yeas he estmated a log cost functon, and found that fo evey pecent se n the numbe of actve contbutos, opeatonal costs se n the ange of.6 to.8 pecent. In sum, thee appea to be lmts to competton n the penson fund management maket. These deve fom the egulaton and fom maket pessues eflectng the pesence of economes of scale. The cost stuctue of penson fund manages and tends Examnaton of the cost stuctue of penson fund manages, and tends ove tme, povdes futhe nfomaton on the behavou and pefomance. Data on the opeatonal costs of the penson fund manages can be obtaned fom the fnancal epots. Opeatonal costs nclude costs of pesonnel (n admnstaton, sales and management), as well as admnstatve, computng and maketng costs, and amotsaton and depecaton. Fgue below, shows the evoluton of opeatonal pe actve contbuto. To facltate compason acoss tme, the sees 5 At ssue s why ae ates of etun and commssons vey smla acoss penson fund manages. Some commentatos ague that egulaton ensues hed behavou among penson fund manages, whle othes blamed pessues fom maket stuctue. 6 Stctly, these ae economes of sze athe than economes of scale. In fnancal nsttutons sze dffeences nomally nvolve moe than a smple scale facto, and nclude output, as well as nput, effcency gans (Bege, Hunte et al, 993). 7

8 ae epoted n constant pces. Total opeatonal costs emaned steady fom 982 to 987, but show a steep change n 988 whch s explaned by the costs of ensuance of suvvo and dsablty beneft labltes. 8 Fom 988 to 992 opeatonal costs se steadly, acceleatng late on due to competton. The ntoducton of stcte egulatons on tansfes n 998 stems the se n costs. Fgue. Opeatonal costs pe contbuto and Total opeatonal costs of penson fund manages Chle Component costs pe contbuto (,s Ch$ Decembe 998) Total costs (mllon Ch$ Decembe 998) Offce pesonnel Sales pesonnel Executve Remuneaton Maketng Costs Compute Costs Admnstaton Costs Total Opeatonal Costs When opeatonal costs ae dsaggegated nto the dffeent components and ae standadsed pe actve contbuto, the sees that the apd se n sales pesonnel costs encouaged by the lage numbe of new entants n 992 s the man facto n the se n opeatonal costs. The stategy was to captue maket shae by attactng contbutos fom exstng penson fund manages, hence the emphass on sales pesonnel. The exstng penson fund manages eacted by expandng the sales foce to potect the maket shae. Fom 994 onwads, the penson fund manages employ moe pesonnel n sales than n admnstaton, and the salay costs of sales pesonnel ae geate than the salay costs of admnstatve and management pesonnel 7 These nclude defned beneft and defned contbuton penson plans. 8 Po to 988, the penson fund manages had not sought to ensue these labltes. 8

9 combned. In addton the numbe of tansfes nceased geneatng exta costs. The new estctons on tansfes ntoduced n 998 ae accompaned by a fall n sales costs as tansfes declne. Oveall opeatonal costs pe contbuto show a tendency to se ove tme. Ths uns counte expectatons that penson fund manages could explot economes of sze, o that competton could foce costs down. The mpact of commsson egulatons Regulaton can educe the scope fo compettve behavou on the pat of penson fund manages. The geneal vew s that egulaton must be stong ntally as the maket s establshed anew, to be elaxed gadually as the maket develops (Vttas 998). Chle's expeence wth egulaton eflects ths appoach. In patcula, egulatons on fund's potfolos have evolved to take account of maket and macoeconomc changes. Nonetheless, some commentatos have agued that egulaton s esponsble fo the mxed pefomance of the penson fund manages. Shah (997) agues that penson fund manages' 'hed' behavou s a consequence of the stct egulaton of penson fund potfolos educng the scope fo competton n asset management, and egulaton on mnmum ates of etun focng fund manages to taget aveage maket etuns. It can also be agued that the poblem s that egulatons ae too loose. Regulatons on commsson and chages ae mpotant n ths context. Penson fund manages chage saves commsson fo the sevces, and ths s the man souce of evenue. The egulatoy famewok defnes what sevces the penson fund manages can chage commsson fo, and the type of commsson, but penson fund manages ae fee to decde whethe o not to mplement these chages, and the level. Sevces on whch chages ae pemtted nclude fund management, etement account deposts, tansfes, deposts and wthdawals fom voluntay savng accounts, wthdawals fom voluntay seveance accounts, and penson beneft payments. The fact that fund manages ae fee to set chages encouaged them to select a commsson stuctue that maxmses the maket advantage vs a vs othe fnancal nsttutons, but one whch s out of lne wth the cost stuctue (Baentos 998). Intally, the penson fund manages chaged fxed and popotonal commsson fo fund mantenance. One effect of ths was to educe damatcally the penson fund of wokes who, due to unemployment o nactvty, 9

10 ceased to contbute egulaly. 9 Advese publc pecepton led to changes n the egulatons. In md 983, nactve contbutos wee exempted fom fxed commsson on the fund balance, and n Januay 988 ths commsson was elmnated fo all. Snce 988, eanngs-elated commsson chaged on deposts has been the manstay of penson fund manages' evenue. Ths s chaged as an addtonal contbuton ove the basc contbuton ate. The penson fund manages have been eluctant to use the full ange of pemtted commssons. In ode to foestall etalaton measues, they have chaged commsson fo tansfes to othe penson fund manages only vey occasonally. They have not chaged fo the management of etement accounts n ode to compete wth nsuance companes offeng penson annutes. They have neve chaged commsson on voluntay savng and seveance savng accounts because ths makes these accounts moe attactve to potental saves elatve to savngs accounts offeed by banks. Ths has poduced a stuaton n whch the full opeatonal costs of fund manages fall upon egula contbutos to ndvdual etement accounts, manly though eanngs elated vaable commsson. As a consequence the commsson stuctue s out of lne wth the cost stuctue, wth advese effects on competton and pefomance. Competton, egulaton and maket stateges Two boad maket stateges have evolved n the penson fund management maket. Some penson fund manages have sought to maxmse maket shae n ode to explot economes of sze n the maket. At the othe end of the spectum, some penson fund manages have focused nstead on attactng hgh eanes. They have chosen to eman elatvely small, and to focus the maketng effots on fndng and etanng hgh eanes. Ths s made ease whee afflates have a stong allegance nutued by tade unons and othe ogansatons. The maket stategy eles on the fact that the ovewhelmng popoton of the fund manages' evenue comes fom eanngs 9 Thee s also concen ove the egessve mpact of fxed commsson as they have a geate mpact upon low eanng wokes, and the accumulated penson fund. The basc contbuton ate s pecent of eanngs, and the addtonal contbuton has fluctuated between 2.4 and 3 pecent of eanngs. The addtonal contbuton ncludes a dsablty and suvvo nsuance pemum. In Decembe 999, fo example, the aveage addtonal contbuton was 2.4 pecent and the nsuance pemum.65 pecent. The net vaable commsson was theefoe.75 pecent of eanngs. In Decembe 999 two penson fund manages, Planvtal and SumaBansande, chaged fxed commsson fo wthdawals unde the phased wthdawal and defeed lfe annuty etement pensons. Two othe fund manages, Cupum and Habtat chaged vaable commsson on the etement fund.

11 elated commsson on etement account deposts. 2 Opeatonal costs, on the othe hand, do not se n lne wth contbutoy eanngs. In fact, to the extent that economes of scale exst n the maket, aveage opeatonal costs ae moe lkely to declne n contbutoy eanngs. Smalle penson fund manages can theefoe absob hgh opeatonal costs by focusng on hgh eanng wokes. A selectve maketng stategy allows small penson fund manages to suvve and pospe n the maket. Competton, f effectve, should esult n the ext of less effcent fms, and the suvval of the moe effcent ones. The poblem s that thee ae two vable stateges obseved n the fund manges' maket, one of whch eles on economes of scale and the othe on attactng the hghest eanes. The dscusson n ths Secton demonstates the complex nteplay of maket pessues and egulaton shapng the behavou of penson fund manages. It also dentfes a numbe of mpotant ssues fo nvestgaton. It s mpotant to evaluate the extent to whch penson fund manages ae effcent. It s also mpotant to consde what ae the key souces of neffcency, and whethe neffcences ae a facto of egulaton o maket stuctue. The length of expeence wth penson efom n Chle povdes the data needed to measue changes n penson fund manage effcency ove tme. The next Secton makes use of data envelopment analyss to thow lght on these ssues. 3. Effcency Measuement and Data Data Envelopment Analyss (DEA) has been used to compae the techncal effcency of elatvely homogeneous sets of poducton unts. It was ntally developed to compae the effcency of publc secto and not-fo-poft poducton unts (Chanes, Coope et al. 978), but moe ecently has been appled successfully to the fnancal secto (Felds and Muphy 989; Fee and Lovell 99; Felds, Muphy et al. 993; Nasse Katb and Matthews 999). It s theefoe appopate to the task of analysng the effcency of penson fund manages n Latn Ameca. These ae estcted by egulaton to povdng specfed and equvalent sevces, and to apply smla commsson and chages to the afflates. 2 Fo the maket as a whole nealy 95 pecent of commsson ncome came fom ths souce n 995.

12 CCR and BCC models In DEA analyss, effcency s defned as the ato of a weghted sum of outputs to a weghted sum of nputs, whee sets of weghts fo each of the obseved poducton unts ae found by solvng the followng model poposed by Chanes, Coope and Rhodes (978) efeed to as the CCR model (Boussofane and Dyson 99) : Max h subjectto t m u y v x j j () t m u v y x j j, j,..., n ; u, v ; and whee n ndexes unts, t ndexes outputs, m ndexes nputs, and y j = amount of output fom unt j, xj = amount of nput to unt j. u = the weght gven to output, v I = the weght gven to nput, = a small postve numbe. In ths model the effcency of unt j s maxmsed subject to the effcences of all unts n the set havng an uppe bound of. The weghts ae teated as unknowns and assumed to be nonnegatve, and wll be chosen to maxmse the effcency of the unt j. The unt obseved s elatvely effcent f the effcency measue h equals and neffcent f less than. The effcency of all othe obseved unts s then measued n tun. Ths model can be conveted to a lnea fom, n ode that lnea pogammng can be appled, as: 2

13 3 Wth ths lneasaton, the denomnato has been set to a constant (abtaly set at ), and the numeato s maxmsed fo each unt n tun. Altenatvely, the dual can be solved, as n: Estmaton of (3) geneates an effcency scoe fo each of the unts obseved elatve to all the othes, whch can be ntepeted as an 'effcent poducton fonte'. The aggegate effcency estmated fo a unt equals the poduct of ts pue techncal and scale effcency. An altenatve model was povded by Banke, Chanes and Coope (984), who extend the CCR model to assess pue techncal and scale effcences of unts. The model takes account of the effects of etuns to scale wthn the unts obseved and dentfes the most poductve scale sze fo each unt. It has the addtonal estcton that the sum of the multples j should add to. The BCC model can be wtten as:.,...,,,,...,,,,...,, (2) m v t u n j x v y u x v subjectto y u Max h m j t j m j t j.,,,,,,...,,,,...,, (3) ned unconsta Z and j s s t y x s m x s Z x to subject s s Z Mn j n j j j j n j j j j m t

14 Mn h subject to hx n j n j j y y j j j j j n s m t s x j, j y j s, s,,..., m,,,..., t, j s, s (4) Data A key ssue n the mplementaton of DEA s the selecton of nputs and outputs. A poducton unt wll use a vaety of esouces to poduce a unt of output of standad qualty. Ideally, one would want to nclude all esouces and all outputs poduced by the fm. In pactce the selecton of nputs and outputs s estcted to those that can be obseved fully, and can be measued wth some accuacy. The selecton of nputs and outputs s also estcted by the sze of the sample of unts obseved. The numbes of nputs and outputs needs to be small compaed to the numbe of unt used n the analyss fo the dscmnatng powes of DEA to be effectve. Boussofane et al (99) suggest that the poduct of the numbe of nputs and outputs s a easonable ndcatos of the mnmum numbe of unts needed to pevent a dstoton of the effcency measue. A futhe ssue wth the selecton of the nputs and outputs efes to the undelyng conceptualsaton of poducton n fnancal sevces. The queston hee s to dentfy what t s that the fnancal povdes poduce The lteatue on applyng DEA to the bankng secto has dentfed two dffeent appoaches (Fee and Lovell 99). One appoach stesses the ntemedaton functon of fnancal povdes, the ntemedaton between saves and boowes. Wthn ths appoach, bankng secto output can be dentfed as, and measued by, the value of loans extended. A second appoach, descbed as the poducton appoach, focuses on bankng sevces and poducts, and measues output by the value of a bank's accounts and tansactons. In some espects penson fund manages esemble banks, but when due account s taken of the 4

15 dedcated natue of etement savng accounts, and the bndng nvestment egulaton, the ntemedaton ole s educed to a mnmum. In the context of penson fund manages, we consde the poducton appoach to be pefeable. Penson fund manages employ a labou foce, use captal equpment and buldngs, and maketng sevces to attact and etan afflates. In vew of ths, the analyss below focuses on two measues of output Total Revenue and the numbe of Contbutos. Eanngs elated contbuton and commsson chages account fo the lagest shae of evenue. Contbutos ae afflates who make egula contbutons, whle afflates ae all those who at any tme have opened a etement savng account. Inputs ae Maketng and Sales Costs, Offce Pesonnel and Executve Pay, and Admnstaton and Computng Costs. As noted above, competton among penson fund manages makes maket shae heavly dependent upon non-pce advantage, that s maketng and sales foce capacty. Ths s captued by the fst nput vaable. The natue of the busness eques admnstatve competence, wth a sgnfcant nfomaton technology component. Ths explans the selecton of the othe two vaables Results and Dscusson The effcency of penson fund manages ove tme Applyng DEA to the data fo penson fund manages fo the peod 982 to 999, esults n a techncal effcency scoe fo each fund manage n the maket n each yea. The aveage effcency scoes fo the maket ove the peod ae pesented n Fgue 2. 3 The data used n the analyss below was extacted fom the balance sheets of the penson fund manages epoted annually to the egulato, and fom publshed penson system nfomaton. (Supentendenca de Admnstadoas de Fondos de Pensones seveal ssues). 5

16 Fgue 2. Fund management maket effcency Chle CCR mean BCC mean CCR mn CCR max CCR weghted mean Outputs: evenue - contbutos Inputs: admn and computng costs - offce and exec.payoll - sales and maketng payoll and costs Data Souce: SAFP Boletn Estadstco, Annual Balance Sheets It wll be useful to focus fst on the mean effcency scoes usng the CCR model. These show thee dstnct phases. The fst, unnng fom 982 to 989, s chaactesed by sng effcency, explaned by the apd ntal gowth n contbutos whch allows the penson fund manages to explot economes of sze. Ths means the lage ntal set up costs ae dstbuted acoss a lage numbe of etement accounts, and thee s a moe effcent utlsaton of captal equpment and banch netwoks. Wheeas at the stat of the maket n 982, penson fund manages could have povded the same sevces wth just ove 4 pecent of the nputs, by 989, neffcency s educed to a thd. The second phase stats n 99 and ends n 994. Ths phase s chaactesed by a maked declne n effcency. At the tough n 994, the penson fund manages etun to the neffcency levels of 982. An mpotant contbutoy facto s the expanson n the numbe of fund manages n the maket and the assocated se n sales and maketng costs. The thd phase begns n 995, wth a steady mpovement n maket effcency acceleatng afte 998. A fall n admnstaton costs afte 994 patly compensates fo the contnued se n sales costs, but t s the pecptous fall n the latte followng estctons on tansfes that explans the apd se n 6

17 effcency n the last yea. The key concluson hee s that thee s no evdence of a sustaned mpovement n the techncal effcency of penson fund manages ove the whole peod. The ange of annual effcency scoes s lage, fom vey poo to vey good pefomes. Agan, thee s no evdence of a tend towads a educton n the ange of effcency scoes of penson fund manages. The dspeson n effcency scoes ses wth the enty of new penson fund manages nto the maket n 992. A weghted aveage of CCR effcency scoes (weghted by the shae of contbutos) s hghe than the unweghted aveage, ndcatng that lage penson fund manages have hghe effcency scoes thoughout the peod. The BCC effcency scoes ae smla to the CCR scoes, but wheeas the effcency scoes obtaned usng the CCR model fal to dstngush between techncal and scale neffcency, but the BCC model takes nto consdeaton scale effects. As a easonable appoxmaton, the dffeence ato of CCR to the BCC effcency scoes povdes a measue of scale neffcences, and theefoe of potental scale effcences. As can be seen fom the Fgue, thee s scope fo scale effcency mpovements among penson fund manages, but techncal ne ffcences ae of much geate mpotance than scale neffcences. The oveall concluson s that the penson fund management maket s neffcent, n the sense that the same sevces could have been povded wth a facton of esouces used. At the end of the peod unde examnaton, penson fund manages could have poduced the same level of output wth aound 35 pecent fewe esouces. Although thee ae changes n effcency ove tme, these do not show a sustaned tend towads contnuous mpovement. It s nteestng that enty of new fund manages esulted n a declne n maket effcency, wheeas a subsequent peod of consoldaton esulted n an mpovement n effcency. The evdence pesented n the Fgue suggests that lage penson fund manages have bette effcency scoes than smalle one. Ths could be explaned by the pesence of sgnfcant scale economes, but the analyss suggests that techncal neffcency s much moe mpotant. 7

18 Penson fund manage effcency and maket stateges Table 2 below shows, fo selected yeas, effcency scoes fo penson fund manages anked n descendng ode accodng to the CCR scoes. It demonstates the wde ange of effcency scoes, wth some vey low scoes. Thee s some pesstence n the ankng of penson fund manages ove tme. Habtat and Cupum, fo example, can be found among the top thee ankngs n all fou yeas. At the othe end of the spectum, t appeas that those penson fund manages wth lowe scoes have a smalle chance of emanng n the maket, especally when compang 994 and 999. The consoldaton n the maket n the late 99s appeas to have elmnated the long tal of fund manages wth low effcency scoes n 994. It s notceable that the fund manages n opeaton n 999 had been, wth a sngle excepton, n the maket fom the vey begnnng. Ths suggests thee ae advantages to ealy entants, whch new entants fnd t dffcult to match. These advantages ae not exclusvely to do wth sze. In fact, as agued above, thee s evdence of two dffeent maket stateges. On the one hand, thee s a goup of lage fund manages who ely on maket shae to explot economes of sze. These tend to spend less pe head of contbuto on sales and maketng because they put less effot on selectng and etanng specfc goups of contbutos. They ae lkely to have lowe spendng on sales pe head of contbuto, and a low contbuto to afflate ato. On the othe hand, othe penson fund manages choose to eman small and focus on attactng and etanng hgh eanes. They have hghe spendng on sales and maketng needed to dentfy hgh eanes, and have hghe ato of contbutos to afflates as hghe eanes ae less lkely to have contbutoy gaps and ae gven ncentves to stay wth the fund manage. The hghe costs faced by the latte goup can be absobed by hghe commsson and chages whch ae manly eanngselated. Relatvely small fund manages wth lowe effcency scoes can thus eman n the maket. Magste s one example, as t has emaned n the maket despte low effcency scoes acoss the selected yeas. 8

19 Table 2. Effcency of penson fund manages AFPs 984 CCR BCC AFPs 989 CCR BCC Habtat Habtat 989 Cupum Povda 989 Summa Cupum Santa Maa Planvtal Concoda Poteccon Magste Summa Povda Futuo El Lbetado Unon Alameda El Lbetado Inveta Inveta Planvtal Santa Maa San Cstobal Concoda Magste AFPs 994 CCR BCC AFPs 999 CCR BCC Cupum Cupum 999 Poteccon Povda Habtat Habtat Summa Planvtal Povda SummaBans Santa Maa Santa Maa Bansande ApotaFomenta Planvtal Magste El Lbetado Magste Concoda Fomenta Unon Futuo Banguada Pevpan Amonza Qualtas Genea Valoa

20 The pesence of these two stateges can be obseved moe clealy n a scatteplot of effcency scoes and the ato of contbutos to afflates on the one hand, and to the sales spendng pe contbuto on the othe. These ae shown n Fgues 3 and 4 espectvely. It can be obseved that some fund manages have hghe than aveage sales spendng pe contbuto, and also hghe than aveage ato of contbutos to afflates. In Fgue 3, Cupum, Apota and Magste n 992 have hghe than aveage ato of contbutos to afflates. Ths ato of contbutos to afflates s lkely to fall ove tme fo all fund manages, as some afflates cease to contbute, but the change s maked fo Magste between 992 and In Fgue 4, sales spendng pe contbuto ses n the peod coveed by the two pont obsevatons, but sales costs ae patculaly hghe fo smalle fund manages, Magste, Apota and Cupum, and elatvely smalle fo lage fund manages such as Habtat, Povda and Santa Maa. The maket segmentaton poduced by these two dffeent maket stateges has tended to declne ove tme because as fewe smalle fund manages eman n busness. 4 Ths s due n pat to ntevenng meges. 2

21 Detemnants of penson fund effcency Is the neffcency of penson fund manages explaned by maket stuctue o by egulaton As dscussed above, some studes ague that the mxed pefomance of penson fund manages s explaned by pesence of economes of sze, and lage nfomaton costs to consumes, whch lmt competton. Othes ague the chef eason fo the pefomance of penson fund manages ases fom the esponses to egulaton. Ths s nvestgated by egessng vaables captung maket stuctue and egulaton effects on effcency scoes. The dependent vaable s the CCR effcency scoe. The ndependent vaables nclude the maket shae of fund manages measued as the facton of total contbutos n a patcula yea (mkshae), whch captues the nfluence of maket stuctue. The mpact of egulaton s poxed by the ato of contbutos to afflates 2

22 (ato), and a measue of sales spendng pe contbuto (sales). The atonale fo usng these vaables n ths context s made clea fom the dscusson n the pevous secton. Less effcent penson fund manages can absob hghe costs f they ae able to select hgh eanes. Fnally, a measue of evenue pe contbuto (evenue) s also ncluded to attempt to contol fo the tend change n these vaables ove tme. 5 The esults of the egesson ae shown n Table 3 below. Table 3. Regesson esults Dependent Vaable: CCR Paamete t-stat Sg. Mean Std. Devaton Constant Mkshae Sales Rato Revenue Mean of LHS = 56.9 ; R 2 =.648 ; Adj.R 2 =.4 ; N= 238 The egesson esults ae vey much as expected. When effcency s measued wth the CRR model, theefoe conflatng techncal and scale neffcences, hghe maket shae s assocated wth hghe effcency scoes, eflectng sze economes. The elastcty of the sze effect s moe than popotonal. The coeffcents assocated wth the vaables poxyng esponses to egulaton have dffeent sgns. Hghe sales spendng pe contbuto s assocated wth lowe effcency scoes, but a hghe ato of contbutos to afflates has a postve effect on effcency scoes. The negatve effect of Sales s lage than the postve effect of Rato, esultng n a negatve net effect. These esults suggest that lage penson fund manages, pehaps because they ae able to take advantage of sze economes, have hghe effcency scoes and contbute to the effcency of the maket; whle fund manages' esponses to the ncentves geneated by egulaton ae detmental to maket effcency. It s lkely that egulaton whch contbutes to the msalgnment of costs and commsson chages s detmental to the effcency of the penson fund management maket. 5 As noted above, the ato of contbutos to afflates, and sales spendng, have a tendency to se ove tme. 22

23 CONCLUSION The pape has nvestgated the effcency of the penson fund management maket n Chle. Penson efom n Latn Ameca has led to the ceaton of a penson fund management maket n nne countes so fa. In the efomed penson systems n Latn Ameca, penson fund manages ae expected to compete to attact and etan afflates, but ae subject to extensve egulaton. The effcency of the new penson povdes wll be key to the success of the efoms, to the mpact of the efoms on the economes of the egon, and to the wellbeng of the futue eldely. Howeve, the pefomance of penson fund manages to date has been mxed. Some suggest ths could have been antcpated gven the lmtatons of competton n fnancal, and patculaly nsuance, makets. The nfomaton equements and costs of consume's decson makng on the one hand, and the pesence of economes of sze on the othe, mpose mpotant lmtatons to competton n these makets. Othes suggest the pefomance of penson fund manages has been unduly ccumscbed by egulaton. Chle povdes a good gound on whch to nvestgate these ssues because t has two decades of expeence wth penson efom, and because t povded the model of egulaton othe countes have followed. The pape began by outlnng tends n the behavou and pefomance of penson fund manages, and dscussed key ssues of competton and egulaton. Whlst competton n the penson fund management maket n Chle s lmted, t was agued that egulaton s ton an mpotant extent esponsble fo the mxed pefomance of fund manages. Some commentatos suggest that the poblem ases because egulaton n the penson fund management maket s too extensve and too stct. A dffeent vew was taken n the pape. It was agued that cuent egulaton allows a msalgnment of costs and commsson chages, makng t possble fo some fund manages to pusue maket stateges whch contbute to maket neffcency. Ths s a poduct of the lmted natue of competton n the penson maket, egulaton, and the ent seekng pactces of both penson fund manages and afflates. Whlst economes of sze mply the aveage costs of povdng sevces n the penson maket declne n contbutoy eanngs, the majoty of commsson evenue comes fom eanngs elated commsson chages. In addton, maketng polcy by the penson fund manages has concentated commsson chages 23

24 on egula contbutos to ndvdual etement accounts, nstead of the wde unvese of uses. Commsson chages theefoe have only an ndect elatonshp to sevce costs These ssues ae nvestgated by applyng DEA to data fo penson fund manages n Chle collected fo the peod The fndngs suggest the penson fund management maket n Chle suffes fom sgnfcant neffcency. Thee ae mpotant changes ove tme, but no tend towads an mpovement n effcency. At the stat of the efom, penson fund manages could have poduced the same output wth 4 pecent of the esouces utlsed. Effcency mpoved dung the 98s, but new entants n the ealy 99s, and the se n sales and maketng costs assocated wth the enty, led to a declne n maket effcency. Regulatons estctng tansfes ntoduced n 998 have succeeded n scalng down sales and maketng costs and mpovng effcency scoes. At the end of the peod, penson fund mana ges could have poduced the same output wth 65 pecent of the esouces utlsed. Compason of effcency scoes fo ndvdual penson fund manages ove tme shows that maket stayes have hghe effcency scoes compaed to maket leaves, thus suggestng that competton woks, albet mpefectly, ove tme, but thee ae mpotant exceptons to ths fndng. Analyss of the detemnants of effcency shows that nceased maket shae contbutes postvely to maket effcency. Sales and maketng costs, on the othe hand, ae detmental to maket effcency. Ths s consstent wth the vew that a key poblem n the opeaton of the penson maket s the msalgnment of commssons chaged wth the costs of the sevces povded. The msalgnment of costs and commsson chages leads to neffcent maket outcomes, wth advese mplcatons fo futue penson outcomes. 24

25 REFERENCES Abuhadba, M. (994), Aspectos Oganzaconales y Competenca en el Sstema Pevsonal, mmeo, Santago. Banke, R. D.; A. Chanes and W. W. Coope (984), Some models fo estmatng techncal and scale neffcences n data envelopment analyss, Management Scence, 3, no. 9, pp Baentos, A. (998), Penson Refom n Latn Ameca, Aldeshot: Ashgate. Baentos, A. (999), The emegng penson fund management maket n Latn Ameca, Jounal of Pensons Management, 5, no., pp Bege, A. N.; W. C. Hunte and S. G. Tmme (993), The Effcency of Fnancal Insttutons, Jounal of Bankng and Fnance, 7, pp Boussofane, A. and R. G. Dyson (99), Appled Data Envelopment Analyss, Euopean Jounal of Opeatonal Reseach, 52, pp. -5. Chanes, A.; W. W. Coope and E. Rhodes (978), Measung the effcency of decson makng unts, Euopean Jounal of Opeatonal Reseach, 2, no. 6, pp Damond, P. (993), Pvatzaton of Socal Secuty : Lessons fom Chle, Wokng Pape 45, Cambdge MA: NBER. Fee, G. D. and C. A. Lovell (99), Measung Cost Effcency n Bankng: Econometc and Lnea Pogammng Evdence, Jounal of Econometcs, 46, pp Felds, J. A. and N. B. Muphy (989), An Analyss of Effcency n the Delvey of Fnancal Sevces: The Case of UK Insuance Agences, Jounal of Fnancal Sevces Reseach, 2, pp Felds, J. A.; N. B. Muphy and D. Ttoglou (993), An Intenatonal Compason of Scale Economes n Bankng, Jounal of Fnancal Sevces Reseach, 7, pp Mesa-Lago, C. and F. Betanou (998), Manual de Economía de la Segudad Socal, Montevdeo: Cento Latnoamecano de Economía Humana. Manda, E. (994), Evdenca de Economías de Escala en las Admnstadoas de Fondos de Pensones, Estudos de Admnstacón,, no., pp Mtchell, O. S. (996), Admnstatve Costs n Publc and Pvate Retement Systems, Wokng Pape 5734, Cambdge, MA: Natonal Bueau of Economc Reseach. 25

26 Nasse Katb, M. and K. Matthews (999), A Non-Paametc Appoach to Effcency Measuement n the Malaysan Bankng Secto, Dscusson Papes n Economcs 99:3, Cadff: Cadff Busness School. Quesse, M. (997), Penson Refom and Pvate Penson Funds n Peu and Colomba, Wokng Pape 853, Washngton DC: Wold Bank. Quesse, M. (998a), Regulaton and supevson of penson funds: Pncples and pactces, Intenatonal Socal Secuty Revew, 5, no. 2/98, pp Quesse, M. (998b), The Second-Geneaton Penson Refoms n Latn Ameca, Wokng Pape AWP 5.4, Pas: OECD. Shah, H. (997), Towads bette egulaton of pvate penson funds, Wokng Pape 79, Washngton DC: The Wold Bank. Snha, T.; F. Matínez and C. Baos-Muñoz (998), Pefomance of Publcly Mandated Pvate Penson Funds n Mexco: Smulatons wth Tansacton Costs, mmeo, Mexco DF: Insttuto Tecnológco Autónomo de Mexco. Supentendenca de Admnstadoas de Fondos de Pensones (seveal ssues), Boletín Estadístco, Santago: SAFP. Valdés-Peto, S. (992), Seleccón de AFP y Regulacón de la Infomacón, Wokng Pape 4, Santago: Insttuto de Economía, Pontfca Unvesdad Católca de Chle. Vttas, D. (998), Regulatoy Contoveses of Pvate Penson Funds, mmeo, Washngton DC: The Wold Bank. Yemo, J. (2), 'Insttutonal Investos n Latn Ameca: Recent Tends and Regulatoy Challenges', n Insttutonal Investos n Latn Ameca OECD (ed.) OECD, Pas, pp