Economies of scale and scope in Australian superannuation (pension) funds Received (in revised form): 11 th September 2012

Transcription

1 Orgnal Artcle Economes of scale and scope n Australan superannuaton (penson) funds Receved (n revsed form): 11 th September 2012 Helen Hggs s a researcher n the Department of Accountng, Fnance and Economcs at Grffth Unversty, Australa. Hggs research nterests nclude the estmaton of cost functons n Australan superannuaton funds. Andrew C. Worthngton s a researcher n the Department of Accountng, Fnance and Economcs at Grffth Unversty, Australa. Worthngton s research nterests nclude performance and consoldaton n the Australan superannuaton ndustry. ABSTRACT Ths artcle provdes estmates of economes of scale and scope for 200 large Australan superannuaton (penson) funds usng a multple-output cost functon. We separately defne costs n terms of nvestment expenses ncludng nvestment, custodan and asset management fees and operatng expenses comprsng management, admnstraton, actuaral, drector and trustee fees / charges. The four nvestment outputs are cash flow-adjusted net assets, the number of nvestment optons, the proporton of total assets n the default strategy and the 5-year rate of return for nvestment costs, whereas the four operatng outputs are cash flow-adjusted net assets, the number of members, net contrbuton flows and net rollovers for operatng costs. We fnd that economes of scale hold up to at least 300 per cent of current mean fund output n both nvestment and operatng costs. There s lttle evdence that economes of scope preval, generally reflected n the proclvty for many superannuaton funds to contract out aspects of both nvestments and operatons. Pensons (2012) 17, do: /pm Keywords: economes of scale ; economes of scope ; cost effcency ; superannuaton (penson) funds INTRODUCTION Australa s superannuaton system (comprsng a compulsory employer-funded superannuaton (retrement ncome) system supplemented by a narrowly focused (age, means and ncome tested) publc penson) contnues to grow apace wth superannuaton savngs forecast to ncrease from US $ 1.1 trllon (90 per cent of GDP) today Correspondence: Andrew C. Worthngton Department of Accountng, Fnance and Economcs, Grffth Unversty, 170 Kessels Road, Nathan, Queensland 111, Australa to $ 6.1 trllon (130 per cent of GDP) by The profle and structure of the superannuaton ndustry has also changed rapdly, wth consoldaton projected to cause the number of large funds to fall from 7 today (73 n 1996) to 7 n 2035, wth the largest fund to ncrease from $ 1.5 bllon today to $ 350 bllon n However, accompanyng these dramatc changes has been dsquet about several lmtatons of the present system n terms of effcency (that t operates n the most cost-effectve manner and n the best nterests of members) and operatons (that returns to 2012 Macmllan Publshers Ltd Pensons Vol. 17,,

2 Economes of scale and scope n Australan superannuaton (penson) funds members are maxmsed, ncludng through mnmsng costs). Reportng n June 2010, the Revew of the Governance, Effcency, Structure and Operaton of Australa s Superannuaton System (Cooper Revew) made a large number of recommendatons. These nclude the substantal benefts for members of ncreased scale n the superannuaton ndustry and the desrablty of the Australan Prudental Regulaton Authorty (APRA) n overseeng and promotng the effcency of the superannuaton system. They also nclude the proposed role of SuperStream (a package of measures amed at reformng back-offce operatons n terms of the ncreased use of technology, unform data standards, tax fle dentfcaton and straght-through processng of transactons) and MySuper (a smple product desgned for the majorty of members) n lowerng overall costs. Clearly, a major focus n these and future developments has been and wll be the level of nvestment and operatng (admnstraton) costs, partcularly ther nteracton wth the scale (the sze of producton) and scope (the dversty of producton) of superannuaton funds. Unfortunately, and outsde the Cooper Revew s partal evdence, there s relatvely lttle work on the role of economes of scale and scope n mprovng the level of cost effcency n superannuaton funds. In fact, just three studes broadly relate to the objectve of ths artcle to estmate economes of scale and scope n Australan superannuaton funds. Frst, Ang and Wuh Ln 2 employ cost functons to estmate economes of scale and scope, but for US mutual funds. Ther emphass s on the parameter estmates for dfferent fund characterstcs (ncludng asset allocaton and nvestment strategy) wth the fndngs ndcatng strong economes of scale n operatng expenses, but only weak evdence of economes of scope n multproduct funds. Second, Malhotra, Marsetty, Vjaya and Arff 3 estmate a translog cost functon for retal superannuaton funds n Australa n 1999 / Puttng asde that ths analyss s now rather dated, the emphass les wth estmates of cost elastctes at the mean and not economes of scale at dfferent levels of output. There s no attenton to economes of scope. Fnally, Bkker and de Dreu,5 formulate models of nvestment and operatng costs for Dutch penson funds, but as wth Malhotra, Marsetty, Vjaya and Arff 3 there s no attempt to estmate economes of scope nor dscuss scale and scope effcences at dfferent levels of observed output. Nevertheless, t s n the sprt of the latter that we frame our analyss. In so dong, we complement the fndng of exstng studes of costs n superannuaton (penson) funds n the Unted States, 6 9 Australa, Ireland 13 and the Unted Kngdom. 1 In the rest of the artcle, we frst brefly dscuss nvestment and operatng costs n superannuaton funds. We then descrbe the specfcaton of superannuaton fund costs and outputs used n the analyss. Afterwards, we descrbe the cost functon used to estmate the economes of scale and scope and present the results. We end the artcle wth a bref concluson. ECONOMIES OF SCALE AND SCOPE IN SUPERANNUATION FUNDS In general, we can dvde the overall costs requred to operate a superannuaton fund nto two areas: nvestment costs and operatng (or admnstraton) costs. To start wth, nvestment-related costs are expenses requred to make and undertake nvestment decsons and, accordng to the Cooper Revew, 1 are qute properly related to the assets under management, prncpally nvestment, custodan and asset management fees. Although we expect the quantum of these costs to ncrease wth assets under management, they wll necessarly vary wth whether the assets are actvely or passvely managed, the nature of the underlyng dstrbuton of assets and the extent to whch the trustee has decded to nvest n alternatve and llqud assets. By comparson, operatng costs n superannuaton funds largely reflect the transacton-based needs of fund members across a wde range of servces, ncludng customer nterface (call centres, Webstes, malrooms, 2012 Macmllan Publshers Ltd Pensons Vol. 17,,

3 Hggs and Worthngton marketng), back offce processng (new members, beneft payments, contrbuton uploads, complance, audt and reportng) and record-keepng systems (nformaton technology, member records, transactons and nvestment support). Importantly, external partes (such as custodan servces) almost unversally handle some of these servces, whereas others are along a spectrum of n-house and external thrd-party provders. Equally mportantly, although most of these costs postvely relate to fund membershp, they have the propensty to declne wth average account sze. In terms of both theory and the goal of maxmsng member benefts net of costs, a possble long-run cost objectve for a superannuaton fund s to be n a poston to produce the desred output (or outputs) at the lowest possble cost (or cost mnmsaton). On the bass of the above dscusson, the prncpal outputs for most superannuaton funds would appear to be assets (for nvestment costs) and members (for operatng costs). Among other thngs, ths means adjustng the scale of producton to the most approprate sze. Sometmes dvdng the producton process nto smaller more specalzed producton unts can result n economes, as evdenced by the allocaton of some superannuaton funds tasks across dvsons or departments. On other occasons, enlargng the scale of producton can acheve lower unt costs. Through ths process, superannuaton funds overcome ndvsbltes n factor nputs, avod the costs of a lack of capacty and gan access to economes n the fxed costs of producton ncludng marketng and admnstraton (ncludng human resources and nformaton technology). Once multple products arse, the presence or absence of complementartes between outputs may provde cost advantages n that a sngle fund producng a gven level of output for a number of product or product dmensons may spend less than a combnaton of several specalsed funds. Wth producton n superannuaton funds, the producton process typcally comprses multproduct attrbutes because t produces multple products (dfferent nvestment optons (IOP), member servces and returns) through the sharng and jont utlsaton of nputs ncludng nvestng management and admnstratve labour, custodan servces, nformaton technology, human resources and so on. SPECIFICATION OF OUTPUTS AND COSTS For our analyss, we use data consstng of fund-level observatons of Australa s 200 largest superannuaton funds. We obtan the data from APRA. Unfortunately, only a sngle cross-secton of data was avalable n Stochastc cost functons typcally regress costs (here nvestment and operatng costs) on the quantty and prce of the factor nputs used n producton (typcally captal and labour) and the outputs produced. Unfortunately, the lmted data released by APRA does not permt full specfcaton of the prces and quanttes of the factor nputs. As these data are not avalable, t amounts to the assumpton that nput prces are constant across the superannuaton ndustry and so the quantty of factor nputs employed n producton s proportonal to the quantty of nvestment expenses. Fortunately, ths s a realstc assumpton n that Australan superannuaton funds are arguably prce-takers operatng n hghly compettve global factor markets. We specfy four outputs each for the nvestment and operatng functons of superannuaton funds. Table 1 provdes selected descrptve statstcs. For nvestment costs these are: cash flow-adjusted assets (CAS) (n $ mllons), the number of IOP, the percentage of funds n defned benefts (DEF) and the percentage 5-year average return (FIV). On ths bass, as the number of IOP ncreases, so too wll the costs assocated wth management, whle a larger proporton of members n a defned beneft fund wll tend to place downward pressure on nvestment costs assocated wth a relatvely more conservatve strategy. Lastly, we expect that hgher nvestment returns wll ultmately only result from ncreased nvestment expenses at the fund level, even f only for the fact that management securng hgher returns wll demand a premum for ts servces Macmllan Publshers Ltd Pensons Vol. 17,,

4 Economes of scale and scope n Australan superannuaton (penson) funds Table 1 : Selected descrptve statstcs CAS ( $ ml.) Number of IOP Percentage of funds n DEF FIV MEM Net CNT ( $ ml.) Net RLL ( $ ml.) Mean Maxmum Mnmum Standard Devaton Skewness Kurtoss Turnng to operatng costs, the four outputs are CAS (n $ mllons), the number of members (MEM), net contrbutons (CNT) (n $ mllons) and net rollovers (RLL) (also n $ mllons). The prncpal argument here s that operatng costs mostly relate to transacton-related servces wthn the fund and that these wll prmarly vary wth the number of fund members. However, they wll also vary wth the average account sze and the frequency and magntude of member transactons wth the fund n terms of contrbutons and rollovers. MODEL SPECIFICATION We use a cost functon to estmate economes of scale and scope n Australan superannuaton funds. The basc assumpton s that superannuaton funds seek to mnmse costs at some chosen level of output. A cost functon that allows the economes of scale to vary wth the dfferent levels of output s: = 1 = 1 C = a + a y + 1/ 2 b ( y ) 0 + d yy + e, j= 1 j j j where α 0 s the fxed cost term, α ( = 1, 2, 3, ) are the slope coeffcents of the lnear term, β ( = 1, 2, 3,) are the slope coeffcents of the quadratc terms, δ j ( = 1, 2, 3,, j = 1, 2, 3, and j ) are the slope coeffcents of the cross-product terms, C s the of the total operatng costs of each superannuaton fund (n $ 000) and y 1 y represent the outputs. The cost functon n (1) allows the estmaton of the economes of scale and scope. These are ray 2 (1) economes of scale, product-specfc economes of scale and product-specfc economes of scope (PES). Under ray economes of scale, we assume that the composton of each superannuaton fund s output remans fxed whle the aggregate sze of output vares. Ths provdes a measure of scale analogous to the sngle output case where ray economes (dseconomes) of scale exst f the measure s greater (less) than unty. Product-specfc economes of scale, however, allow one output to vary, whle others are constant. Product-specfc economes (dseconomes) of scale then exst f the measure s greater (less) than unty. Fnally, PES measures whether the cost of producng the outputs jontly s less than the costs of producng them separately. A value greater than or equal to 0 thus ndcates that cost advantages accrue through the jont producton of outputs. The method for calculatng these measures s as follows. Frst, the average ncremental cost, AIC ( y ) for producng output y s defned as: Cy ( ) Cy ( N ) AIC( y ) =, y = 123,,, where C ( y ) s the total cost of producng the outputs and C ( y N 1 ) s the total cost of producng 0 unts of the th output. In the case of a sngle product, we measure the economes of scale usng the AIC dvded by the margnal cost (MC). The product-specfc economes of scale for y, E ( y ) are specfed as: AIC( y ) Ey ( ) = MC( y ) (2) (3) 2012 Macmllan Publshers Ltd Pensons Vol. 17,,

5 Hggs and Worthngton where MC ( y ) = C / y s the MC of producng y unts of output. Ray economes of scale exst when the quanttes of the product ncrease proportonately: E( RAY ) = = 1 y Cy ( ) MC( y ) If E ( y ) or E ( RAY ) s greater than 1 (less) then economes of scale (dseconomes of scale) exsts for output y. Second, we can dvde economes of scope nto global economes of scope ( GES ) and PES and these are: The PES are: GES( y ) = = 1 Cy ( ) Cy ( ) Cy ( ) Cy ( ) + Cy ( N ) Cy ( ) PES( y ) = Cy ( ) EMPIRICAL RESULTS Table 2 presents the estmated coeffcents, standard errors and p -values of the nvestment and operatng cost functons. The R 2 for the nvestment cost functon reported on the () (5) (6) left-hand sde of Table 2 s 0.670, whereas that for the operatng cost functon on the rght-hand sde s We test the null hypotheses of no output effects jontly n addton to varous tests of no lnear, quadratc and output cross-product effects wth χ 2 test statstcs. We reject all hypotheses at 1 per cent level of sgnfcance, thus suggestng that all explanatory varables should jontly be ncluded n estmatng the nvestment and operatng cost functons for Australan superannuaton funds. We use the estmates n Table 2 to calculate the MC and AIC for each of the four superannuaton fund outputs (CAS, IOP, DEF and FIV for nvestment costs and CAS, MEM, CNT and RLL for operatng costs). As shown n Table 3, we do ths for levels of mean output runnng from 50 per cent to 300 per cent (that s, 100 per cent s the mean output n the sample data). The mean values are $ 335 for CAS, 97.1 for IOP, 51.3 per cent for DEF, 3.37 per cent for FIV, for MEM, $ for CNT and $ for RLL. Consder the MC of nvestment on the left-hand sde of Table 3 n the lower panel. As shown, nvestment costs declne contnuously as output (as measured by CAS) ncreases from 50 per cent to 300 per cent of the current mean output of Australa s 200 largest superannuaton funds. Ths s also the case for the number of IOP suggestng that larger funds wth more IOP fnd Table 2 : Estmated cost functon Investment costs Varable Coeffcent Standard Error Operatng costs Probablty Varable Coeffcent Standard Error Probablty C α C α CAS α CAS α IOP α MEM α DEF α CNT α FIV α RLL α CAS 2 β CAS 2 β IOP 2 β MEM 2 β DEF 2 β CNT 2 β FIV 2 β RLL 2 β CAS IOP δ CAS MEM δ CAS DEF δ CAS CNT δ CAS FIV δ CAS RLL δ IOP DEF δ MEM CNT δ IOP FIV δ MEM RLL δ DEF FIV δ CNT RLL δ Macmllan Publshers Ltd Pensons Vol. 17,,

6 Economes of scale and scope n Australan superannuaton (penson) funds Table 3 : Margnal and average ncremental costs Level Investment costs Operatng costs CAS IOP DEF FIV CAS MEM CNT RLL AIC 50 % % % % % % % % % % MC 50 % % % % % % % % % % t ncreasngly less costly on a margnal bass. However, MC ncrease as the share of DEF n the fund and the FIV ncreases, suggestng that better nvestment performance comes at a prce to net benefts. A smlar pcture emerges wth the operatng cost functon on the rght-hand sde of Table 3. As shown, the man drver of operatng costs at hgher levels of output appears to be sze of the fund n terms of CAS and the amount of fund rollovers (RLL). At all but the lowest level of mean output, MEM and the level of contrbutons (CNT) have only a small mpact on operatng costs, suggestng that most operatng costs are farly small per member and dollar value of contrbutons once some mnmum fund sze s attaned. The upper panel of Table presents the product-specfc (E) and ray (RAY) economes of scale for nvestment and operatng costs. As defned, the pont estmates represent the ray economes (dseconomes) of scale: f the pont estmate s greater than unty then ray economes of scale exst across the outputs. As shown, ray economes (the proportonal augmentaton of output holdng composton constant) exst from 50 per cent to 300 per cent of the mean output over the sample perod for both nvestment and operatng expenses. Ths suggests that the sector as a whole s currently experencng economes of scale and there s a clear ncentve to expand the producton of all outputs to explot exstng potental scale economes. However, whereas the ray economes for nvestment costs are contnuously ncreasng at an ncreasng rate, those for operatng costs are ncreasng at a decreasng rate. Ultmately, at some level above 300 per cent of the current mean output, dseconomes of scale wll nfluence operatng costs n Australa s superannuaton funds. The upper panel n Table also ncludes the product-specfc economes of scale. These are the scale economes that exst where an output ncreases n solaton. As shown, for nvestment costs, there are product-specfc economes of scale from 150 per cent to 300 per cent of current mean output for CAS, from 100 per cent to 300 per cent for IOP and from 250 per cent to 300 per cent for FIV. Ths complements evdence from the prevous table that the man drvers of scale economes n superannuaton funds are ncreasng the level of assets under management and the number of IOP. Turnng to 2012 Macmllan Publshers Ltd Pensons Vol. 17,,

7 Hggs and Worthngton Table : Economes of scale and scope Level Investment costs Operatng costs E (CAS) E (IOP) E (DEF) E (FIV) E (RAY) E (CAS) E (MEM) E (CNT) E(RLL) E(RAY) 50 % % % % % % % % % % PES (CAS) PES (IOP) PES (DEF) PES (FIV) GES PES (CAS) PES (MEM) PES (CNT) PES (RLL) GES 50 % % % % % % % % % % Notes : Arrows ndcate the range over whch economes of scale (scope) are found. operatng costs, the product-specfc economes of scale only operate for relatvely narrow bands at low levels of output. Ths suggests that scale economes n operatng costs can only be realsed when ncreasng all transacton-related actvtes n proporton wthn the same fund. Unlke many other ndustres, producton n Australan superannuaton funds does not typcally nvolve full jont producton of the varous nvestment and operatng outputs gven the tendency for funds to choose varous combnatons of contracted-out and n-house provson of both nvestments and operatons. Ths would appear justfed n the lower panel of Table where there are GES only for operatng costs at hgh levels of mean output (150 per cent 300 per cent). There would appear to be no economes of scope possble n nvestments overall, and only some PES for the IOP, DEF and FIV at extremely low levels of output. Overall, ths would suggest that there s lttle ncentve for superannuaton funds to undertake all of ther varous nvestment and operatng actvtes n-house once the fund has reached just 50 per cent 75 per cent of the current level of mean output. Beyond ths pont, there are certanly strong economes of scale n both nvestments and operatons, and thus an ncentve for the fund to grow larger through ncreasng assets under management and the MEM through growth or fund consoldaton, but lttle evdence of economes of scope. Hence, t s hghly cost-effectve for large superannuaton funds to contract-out many of ther nvestment and operatng actvtes. CONCLUSION Ths artcle dscusses the results of a study of economes of scale and scope n Australan superannuaton fund. The man fndngs are as follows. Frst, there s an evdence of ray economes of scale (assumng that the composton of outputs remans unchanged) up to at least 300 per cent of current mean output n both nvestments and operatons. After ths pont, there s some evdence that Macmllan Publshers Ltd Pensons Vol. 17,,

8 Economes of scale and scope n Australan superannuaton (penson) funds dseconomes of scale n operatons may arse and so there s lttle ncentve for superannuaton funds to expand output far beyond that level. However, the economes of scale n nvestment costs ncrease at an ncreasng rate up to 300 per cent of the current level of output, and ths suggests that dseconomes n nvestment costs (f they do exst) wll only apply at very much hgher levels of output than currently found n the Australan superannuaton fund ndustry. Second, n contrast, the evdence for economes for scope s very weak, wth GES only found n operatons and then only at very hgh levels of output. Ths suggests that only the largest funds wll beneft n cost terms from full n-house producton of admnstratve servces and that for the typcal superannuaton fund there are substantal cost savngs n contractng-out many admnstratve tasks and nearly all nvestment actvtes (or at least there are no sgnfcant cost savngs for n-house producton). REFERENCES 1 Revew of the Governance, Effcency, Structure and Operaton of Australa s Superannuaton System (Cooper Revew). ( 2010 ) Super System Revew, Fnal Report Part One: Overvew and Recommendatons. Commonwealth of Australa, Canberra, accessed 15 August Ang, J. S. and Wuh Ln, J. ( 2001 ) A fundamental approach to estmatng economes of scale and scope of fnancal products: The case of mutual funds. Revew of Quanttatve Fnance and Accountng 16 (3) : Malhotra, D. K. K., Marsetty, V. B. and Arff, M. ( 2001 ) Economcs of scale n the retal superannuaton funds n Australa. EFMA 2002 London Meetngs. Avalable at SSRN, accessed 15 August Bkker, J. and de Dreu, J. ( 2006 ) Penson Fund Effcency: The Impact of Scale, Governance and Plan Desgn. DNB Workng Paper no. 109, 5 Bkker, J. and De Dreu, J. ( 2009 ) Operatng costs of penson funds: The mpact of scale, governance and plan desgn. Journal of Pensons, Economcs, and Fnance 8 (1) : Mtchell, O. S. and Andrews, E. S. ( 1981 ) Scale economes n prvate mult-employer penson systems. Industral and Labour Relatons Revew 3 () : Mtchell, O. S. ( 1998 ) Admnstratve costs of publc and prvate penson plans. In: M. Feldsten (ed.) Prvatzng Socal Securty. Chcago, IL: Unversty of Chcago Press, pp Ghlarducc, T. and Terry, K. ( 1999 ) Scale economes n unon penson plan admnstraton: Industral Relatons 38 (1) : Barber, B., Odean, T. and Zheng, L. ( 2005 ) Out of sght, out of mnd: The effects of expenses on mutual fund flows. Journal of Busness 78 (6) : Clare, R. ( 2001 ) Are admnstraton and nvestment costs n the Australan superannuaton ndustry too hgh?. Assocaton of Superannuaton Funds of Australa, asn.au, accessed 15 August Bateman, H. ( 2001 ) Dsclosure of Superannuaton Fees and Charges, Centre for Pensons and Superannuaton. Dscusson Paper no. 03/0, 12 Bateman, H. and Mtchell, O. S. ( 200 ) New evdence on penson plan desgn and admnstratve expenses: The Australan experence. Journal of Penson Economcs and Fnance 3 (1) : Mahon, A. ( 2006 ) Irsh occupatonal pensons: An overvew and analyss of scale economes. Presentaton to the Irsh Assocaton of Penson Funds, 1 Audt Commsson. ( 2006 ) Effcency Challenge: Cost of Admnsterng Local Government Penson Funds n London. Local Government Natonal Report, Macmllan Publshers Ltd Pensons Vol. 17,,