C S 03/03. New Evidence on Pension Plan Design and Administrative Expenses: The Australian Experience. Olivia Mitchell and Hazel Bateman

Transcription

1 P Discussion Paper 03/03 Centre for Pensions and Superannuation C S New Evidence on Pension Plan Design and Administrative Expenses: The Australian Experience Olivia Mitchell and Hazel Bateman

2 New Evidence on Pension Plan Design and Administrative Expenses: The Australian Experience Hazel Bateman and Olivia S. Mitchell* April 9, 2003 Abstract Policymakers seeking to design efficient and smoothly-functioning pension systems for their aging workforces are beginning to acknowledge the key importance of administrative expenses when formulating rules for pension plan structure and fee disclosure requirements. This study explores the links between retirement plan offerings and pension expenses for a wide range of private and public sector pension plan types, using an invaluable new data set on two thousand Australian pension funds. Our analysis indicates how pension plan design can strongly influence plan expenses and consequently eventual retirement security. Hazel Bateman School of Economics, University of New South Wales Sydney 2052, Australia H.Bateman@unsw.edu.au Olivia S. Mitchell Professor of Insurance and Risk Management The Wharton School, 3641 Locust Walk, Rm. 304 CPC Philadelphia PA, T: 215/ h mitchelo@wharton.upenn.edu *The authors are grateful for data provided by the Australian Prudential Regulatory Authority under restricted access conditions. Research support was provided by the Economic Research Institute, the Wharton -SMU Research Center at Singapore Management University, the Australian Research Council, the National Bureau of Economic Research, and the Pension Research Council. Matthew Williams provided able research assistance. Helpful comments were provided by the Editor, Atsushi Seike, Noriyuki Takayama, and Alex Dunnin, though any errors or opinions remain the authors responsibility. This study is part of the NBER program on the Economics of Aging. Bateman and Mitchell, All Rights Reserved.

3 New Evidence on Pension Plan Design and Administrative Expenses: The Australian Experience Hazel Bateman and Olivia S. Mitchell Policymakers in developed nations have become increasingly aware of the need to restructure their national retirement systems, to meet expectations for their rapidly aging populations. Substantive pension reforms have been launched in dozens of countries, frequently involving conversion of a traditional pay-as-you-go system into a funded pension arrangement (see Table 1). Even Germany, Japan, and the United States, countries slow to undertake pension reform, have acknowledged that new approaches are needed to support a secure retirement. For example, the President s Commission to Strengthen Social Security in the U.S. recently recommended that workers be allowed to redirect part of their payroll taxes into funded individually controlled pension accounts (Cogan and Mitchell, 2003). Table 1 here In the process of restructuring retirement systems, reformers recognize that pension structures must be carefully fashioned to ensure that workers deferred earnings are invested judiciously and managed prudently. A crucial issue, of course, is the net rate of return delivered by the pension plan, as this helps determine how high benefits will be during retirement. But pension performance, in turn, relies on the level and structure of administrative charges levied on the pension assets during the accumulation phase. To understand how administrative charges can erode a worker s pension savings, imagine Plan A that deducts one percent of participant assets in administrative fees annually over a 40-year work horizon, and Plan B deducts two percent of assets over the same horizon. While these fee differences might seem small, they add up: the retiree under Plan A will have accumulated 27% more pension plan assets by retirement age, compared to his counterpart in Plan B (Bateman et al., 2001). Nonetheless, plan participants are often unable to assess how small differences in pension administrative charges can yield substantially different retirement benefits, and little is known about how plan design translates into cost structures conducive to higher returns.

4 2 In the present study, we offer new insight into the links between administrative charges and pension design, by using a unique dataset from Australia that ties plan structure to costs in a rich sample of pension plans of varying size, benefit type and sponsor arrangements. In contrast to the U.S., the Australian pension environment permits considerable diversity of plan design, structure, and size, thus enabling a researcher to explore cost issues using much greater cross-sectional variability than available elsewhere. For example, some pensions in Australia are strictly privately managed, others are public funds, still others are employer-sponsored, and some are retail plans, sold to the general public. There are also single-employer and industrywide pension plans, mandatory and voluntary plans, and pensions of both the defined benefit and defined contribution types. These pension plans are also distinguished by operational characteristics, including the degree of member choice of plan they permit, and what investment strategies they pursue. Drawing on this rich information, our analysis uncovers important differences in pension plan expenses resulting from plan design and size, and further it suggests which types of market structure are most cost-effective in converting retirement saving into retirement wealth. In view of the global trend toward funded pension accounts, our analysis has important implications for international pension policy design. Mandatory Private Pensions under the Australian System In 1992, Australia adopted a national mandatory pension law requiring all employers to offer retirement saving plans under a program known as the Superannuation Guarantee (SG). Voluntary occupational retirement saving plans had previously been in existence but they covered only one-third of private and two-thirds of public sector employees. Under the mandatory SG system, all employers must now contribute 9% of payroll on behalf of their employees to a funded pension program (Bateman et al., 2001). Normally the employer is the plan sponsor and decides on his own pension plan structure and plan features, including type of plan, investment patterns, governance rules, and other elements. The employer can elect to manage the plan assets in-house or can outsource them, he may participate in an industry-wide

5 3 plan, or he can allow workers to contract-out their retirement saving by participating in a retirement program offered at the retail level. All pension plans with five or more members are regulated by and must submit annual reports to the pension industry regulator, the Australian Prudential Regulatory Authority (APRA), indicating aspects of plan structure and performance. 1 These annual reports, filed electronically, are a high-quality source of information on plan and sponsor type, assets under management and membership, details on plan governance, asset allocation, outsourcing, and numerous other operational features. Administrative expenses are reported under a heading all expenses of an administrative and investment management nature. 2 These annual returns must be certified by an approved auditor and signed by plan trustees; compliance is close to 100%. Compared to the U.S., which requires plan sponsors to file Form 5500, the Australian data are far more timely and more complete. To examine links between administrative expenses and plan structure, we have been granted access to the complete file of annual returns of Australian superannuation plans under APRA s jurisdiction, accounting for virtually all retirement system members and assets. 3 In the analysis to follow, we concentrate on pension plans with at least 10 members and at least $A1 million in assets, which results in a sample of 1,920 plans representing 80% of superannuation assets and plan members under APRA s jurisdiction. The particular value of this dataset is its inclusion of a wide range of plan types, plan structures, and plan sponsor formats. For the present purposes, we group explanatory variables of most interest into the following categories: Plan scale: measured by assets and membership; Plan type: specified as defined benefit or defined contribution; Sponsor type: indicating whether the plan is offered by an employer or sold in the retail market. Employer-sponsored plans may be offered by a single company in the private or public sector, or these can be multi-employer plans 1 Prior to 2000, small self managed plans with fewer than five members were regulated by the Australian Taxation Office; since 2000 these plans have been allowed to elect to be regulated by APRA instead, and some 8,000 have chosen to do so. 2 APRA annual return forms are available at 3 The data set was made available only under stringent security conditions and all individual identifiers have been removed from the file so no individual superannuation plan can be identified.

6 4 that pool participants within a given industry. 4 By contrast, a retail plan is open to the general public; Operational features: indicating aspects of the plans relating to collection of contributions, account management, asset management, payment of benefits, and reporting and disclosure. In the analysis sample, just over half the plans (57%) are of the defined contribution variety, with the remainder being defined benefit (or hybrid; see Table 2). This is not surprising since the Superannuation Guarantee regulation, specified in terms of a mandatory contribution, somewhat favors the establishment of defined contribution (DC) plans. In certain circumstances, defined benefit (DB) plans may be approved but they are subject to more onerous regulations and actuarial certification. In the singleemployer private market, the DC/DB split is around even, whereas defined contribution plans predominate in the multi-employer market and in the retail sector. It is also interesting to note that almost one-third of the public sector plans in Australia are of the DC type, 5 a pattern that diverges markedly from the US public pension arena where the DB model continues to dominate (Mitchell and Hustead, 2000). In the sample, 80% of plans are sponsored by a single employer, 6% by multiple employers, and 7% are offered in the retail market; 83% of plans are closed while 17% are open. Table 2 here Turning to expense differences across plans, Table 3 shows that reported administrative expenses per member were about $A560 per member per year (about $US320); 6 this figure is in the mid-range of the comparable US data for defined benefit plan expenses ($US70-600). The mean reported expense ratio is 1.08 percent of assets, or 108 basis points (bp). This is much higher than in the US, where plan expenses as a fraction of assets are between bp. There is also substantial variation in the Australian data, as revealed by the very long right tail in Figures 1 and 2. 4 In Australia private sector plans with single employer sponsors are called corporate funds, public sector plans with single employer sponsors are referred to as public sector funds and multi employer sponsored plans are called industry funds. 5 This results partly from the fact that the Superannuation Guarantee explicitly applied to public sector workers previously uncovered by pension systems. In addition, growing awareness of large liabilities under unfunded defined benefit plans has prompted public sector employers to close them down over time, replacing them with defined contribution plans. 6 We assume an exchange rate of $A1 = $US.60, and all data are in 1999 nominal terms.

7 5 Though most plans cost less than 1% of assets (100 bp) per year, some cost up to 400 bp, a rate that can have a potent effect on retirement accumulations. Table 3 and Figures 1 and 2 here We hypothesize that DB plans would be more expensive to offer than DC pensions, due to the higher costs associated with hiring actuaries, reserving for specified benefit promises, and paying guaranteed annuities in the former plantype. Data in Table 3 confirm this view, with DB plan expenses one-third higher (35%) than the sample mean, while DC plans are 25% less expensive. To put it differently, reported DB plan participant costs were 85% higher than those reported by DC plans. However because DB plans are larger, on average, their costs as a percent of assets are lower, tallying only 68 basis points instead of 93 basis points at the median. Below we hold constant asset size to determine whether DC plans maintain their cost per member advantage in a multivariate setting. We also hypothesize that open plans particularly retail may be more expensive to administer in view of their need to advertise and compete in the marketplace, as compared to closed plans not subject to such pressures. 7 Working in the other direction, however, is the fact that retail plans face competitive pressure on fees and service which could render them less expensive in practice. Which effect is stronger is an empirical question. The evidence in Table 3 suggests that retail plans had somewhat lower reported expense levels than employer-sponsored plans. This simple tabulation may be misleading, however, since retail funds are relatively immature in Australia, having smaller asset pools as compared to long-standing employersponsored plans in both the public and private sector. This hypothesis is plausible since costs as a fraction of assets are higher for retail than employer-based pensions. Below we also investigate whether retail plans continue to have lower expenses, after controlling for asset size. Our initial analysis also uncovered some interesting differences within the employer-sponsor category, in that multi-employer plans reported annual per-member expenses much below those for single-employer plans (results not tabulated here). This 7 While most open plans are retail plans, a number of multi employer plans have opted to become open in order to expand their membership to the self-employed working in their industry.

8 6 might indicate that multi-employer plans perform fewer services than do other plans, or alternatively they might underreport their expenses. In the US, however, multi-employer plans tend to report higher expenses than single-employer pensions because the multis have more complete expense information reported to them than do single-employer plans. This is as a result of single employer plans handling expenses for payroll and other related benefits internally rather than computing and charging all relevant expenses to the pension plan. Multivariate Methodology and Results Previous research examining pension expenses in a multivariate setting has employed a cost function approach to ask how U.S. pension expenses relate to money management, record-keeping, and other practices associated with collecting the contributions, managing the plans, and paying benefits. 8 Adapting this approach to the more diverse Australian pension environment allows us to test whether scale economies are evident in the Australian case, since production inputs tend to be lumpy investments (e.g. computers, payroll systems, money management systems, and accounting systems). Hence as the pension system expands, plan expenses could rise less than proportionately as the plan takes advantage of these scale economies. If the evidence confirms that there are substantial cost savings from larger pension plans, this would imply that merging and coalescing groups of participants and plan assets into larger pools could save participants and plan sponsors money. Conversely, breaking up large pension pools might make retirement systems less cost-effective, rather than more so. In relation to pension scale, we hypothesize that pension plan expenses will be differentiated by plan design, and that some pension plan structures will be more efficient than others in converting contributions to retirement incomes. Given the global trend away from single employer-sponsored defined benefit plans towards defined contributions plans offered in the retail sector, we are particularly interested in the expenses of employer-sponsored versus retail plans and defined benefit versus defined contribution plans. To test for scale economies in the Australian context and investigate 8 See Hsin and Mitchell (1997), Mitchell and Hsin (1997), Mitchell (1998), and James et al. (1999).

9 7 the impact of plan design on administrative expenses, we estimate a multiproduct cost function of the following form with the data described earlier: 9 ln(ci) = α0 + α1ln(parti) + α2ln(assetsi) + α3 X + εi (1) where Ci is the administrative expenses of the pension plan, ln(part i ) and ln(assets i ) refer to the (natural log of ) participants and plan assets, and X refers to other plan design factors as described below. 10 Specifically, we test whether α 1 and α 2 are each less than 1, which if so would indicate evidence of scale economies. That is, expenditures for investment management are expected to rise with assets, but if scale economies exist, a plan with more assets will incur lower administrative expenses per dollar invested. Similarly, administrative expenditures would be expected to rise with participants, but if economies of scale exist, growth in the number of participants would generate less than a proportionate increase in administrative expenses. Further, if DB plans are more costly to manage than are DC plans, and retail plans in a decentralized environment are more costly than those in a more centralized environment, the coefficients on those variables will be expected to be positive as well (Mitchell 1998, James et al., 2001). Other design features are also included such as pension plan type (expressed as a 0/1 qualitative variable with defined contribution being the reference category), and sponsor type (employer-sponsor is the reference group). Control variables include several operational factors such as who pays pension contributions (member contributions as a proportion of total contributions, inward transfers as a proportion of assets), indicators of account management (in-house account management is the reference group), indicators of asset management (fraction of assets managed externally), how benefits are paid (benefits paid as a proportion of assets, outward transfers as a proportion of assets), and approaches to reporting and disclosure (no external review is the reference group). The cost model implies that having more assets and participants will add to pension plan administrative expenses, but at a less than one-for-one rate if scale 9 This approach follows Mitchell and Andrews (1981); a range of other functional forms was tested but the fit was inferior to results reported here. 10 As in most cross-section cost function studies, input prices are not explicitly included in the analysis since salaries for actuaries, lawyers, accountants, and financial advisors are set in a nationally competitive market and should not vary systematically across pension systems.

10 8 economies exist. That is, coefficients on these two variables are expected to be positive but less than one if the hypothesis is supported. Results in Table 4 are consistent with this hypothesis: that is, having more assets and participants implies higher administrative expenses, but there are clear signs of scale economies. This is reflected in the Model 1 c oefficient on (ln) assets of around 0.5, indicating that a onepercentage point change in assets raises costs by about half a percentage point (holding participants constant). The participant elasticity is smaller, indicating that a one-percentage point growth in participants (holding assets constant) raises costs by 0.4 percent, again evidence of scale economies. The asset coefficient is comparable to that measured in previous studies for the US public sector, and both coefficients are larger than in the US private sector context. The Australian participant service coefficients are only half those found in US private sector pensions, suggesting more substantial economies of scale along this dimension. Table 4 here Results for other explanatory variables are also interesting. First, defined benefit plans are detected to have one-third higher reported expenses than defined contribution pensions, even after controlling on plan size. This conforms with earlier tabular results. By contrast, retail plans have 70% higher expenses than plans with employer sponsors, after controlling on plan size. 11 This is important since there is a strong political move in Australia to let workers contract out of their employer pensions, a move which would have consequent cost impacts according to our results. Among the operational variables, we find a positive and significant relationship between expenses and benefit payments as well as outward transfers, and a negative relationship between member contributions and inward transfers. It is worth noting that outsourcing of account or asset management does not raise expenses over in-house management, nor does offering member choice. To further investigate how plan design influences administrative costs, we extend Model 1 in two ways. One effort interacts all variables with the DB plan type dummies (Model 2), to determine what in particular is driving the higher DB costs. The second

11 9 effort seeks to identify the impact of sponsor type, so all variables are interacted with retail dummies (Model 3). The results imply that the findings regarding scale economies are robust, and most of the interactions are statistically insignificant. 12 Estimated responses summarized in Table 5 indicate predicted administrative expenses per member and as a percent of assets, holding constant other variables at alternative values typical of small, median, and large pension plans. For small plans, an employersponsored defined contribution model appears to be the lowest-cost design, with absolute dollar levels at $A307 per annum, while the retail plan design would be the highest-cost, with an annual cost of $A620. For small plans the same rankings apply regarding asset-based fees: an employer-sponsored defined contribution model is least costly to operate, with a small plan spending about 77 basis points in expenses (or 0.77% of assets) annually, while the employer-sponsored defined benefit model costs 107 basis points and the retail plan costs 155 basis points. Plan expenses are less strongly differentiated as pension plans grow larger. For instance, a large retail plan would cost an annual $A193 per member (or 48 bp), versus $A95 per member (24 bp) for a large employer-sponsored DC plan or $A134 per member (33 basis points) for a large employer-sponsored DB plans. Table 5 here At the end of the day, what impact would such costs have on workers retirement saving and how would these vary by pension plan size and type? We compute that a worker in a small plan faces a substantial impact of high pension costs. That is, predicted administrative expenses would depress his retirement accumulations by 15% in an employer-sponsored DC plan, which is probably the best he can do, versus 21% for an employer-sponsored DB plan, and 28% in a retail plan. As anticipated, the impact of costs shrinks as the plan grows larger but the effect differs by plan type. Thus a worker s retirement asset is diminished by only 5% in large employer-sponsored DC 11 The impact of alternative employer sponsors was tested, but after controlling for plan size and type, there was no statistically significant difference in administrative expenses between single-employer, public sector, or multi employer funds. 12 An F-test rejects the hypotheses that the entire vector of interactions is equal to zero (for both models); this is mainly because allowing member contributions raises expenses for defined benefit plans (as compared to defined contribution plans) but lowers expenses for retail plans. In retail plans, it is less costly to manage benefit payouts and expenses rise less with participants than for employer sponsored plans; on the other hand, expenses rise with assets faster in the retail than the employer sponsored plans.

12 10 plans, and 7% for employer-sponsored DB plans. For the median Australian pension plan, administrative expenses would be predicted to cut retirement accumulations by 12% in the employer-sponsored DC case, 16% for the employer-sponsored DB case, and 22% for the retail plan case. 13 Discussion Information regarding plan design and pension administrative costs should be of keen interest to those seeking to build pension security. Our analysis of the Australian pension industry should be of general interest because of the wide range of plan design models that have developed in the context of that country s mandatory employersponsored retirement saving system. These findings may be compared to other studies investigating charges levied to customers of other financial products. James et al. (1999) explored charges levied by US retail mutual funds; Murthi et al. (2001) analyze charges associated with UK voluntary pension plans; and Whitehouse (2000) and Bateman et al. (2001) survey administrative charges in Latin American pension systems. The key difference between the present study and previous analyses is that prior research focuses only on charges levied on plan customers, rather than on expenses reported by pension plan managers, and other studies do not capture the diversity of plan types and features that we have explored. Our findings may be summarized as follows: As is true in the US, larger pension plans in Australia are more costly to manage than smaller ones, but expenses rise less than proportionately. Mean pension expenses in the Australian system average about $US 320 per member per year or 1.08 percent of assets. While the dollar figure is not far from the US range, the expense ratio is far higher. The asset effect is similar to that measured in the US public sector pensions plans, but larger than in US private sector pension plans. The participant effect is only half that found for US private sector plans, suggesting substantial economies of scale in the Australian context. Rich data on a wide array of pension plan structures shows that the least costly Australian pension design is an employer-sponsored DC plan. DB and retail pensions available to the general public are 30-70% more costly. 13 These values are computed assuming a 9% contribution (as per Australian SG law) over 35 years with 1% real wage growth and 5% real rate of return; administrative expenses are taken from Table 5. It is assumed that administrative exp enses fully translate into fees borne by the plan member.

13 11 Pension expenses, when incurred over a worker s lifetime in the labor market, can substantially erode the retirement assets. The impact is most striking for workers covered by small plans and retail plans. Some policy questions we hope to address in future work flow from this analysis. For instance, one might ask whether some plans are simply too small to be costeffective under a mandatory pension system. Another question is whether workers might be permitted to manage their pension saving using retail mutual funds, instead of restricting them to employer-sponsored pensions: this change is in the works in Australia, despite the additional costs associated with retail plans, and the Bush administration is presently seeking to encourage retail retirement saving plans. A related issue is that DC plan sponsors are under pressure to offer increased choice over pension investments; this has been on the political agenda in Australia for several years and choice is being implemented in Chile, Mexico, and Singapore among other countries. Such a move raises questions about what should be disclosed to employee participants regarding fees, since workers who fail to understand how expenses vary with plan design may end up with inadequate retirement income. Perhaps a way to rectify this knowledge shortfall would be to allow employee representation on pension boards, and in future research we hope to link fund governance measures with costs and investment performance.

14 12 References Australian Prudential Regulatory Authority (APRA), Annual Report, 2002, APRA, Sydney. Australian Prudential Regulatory Authority (APRA), Superannuation Trends, June Quarter 2002, APRA, Sydney. Bateman, Hazel, Geoffrey Kingston and John Piggott. Forced Saving: Mandating Private Retirement Incomes, Cambridge University Press Clark, Robert L. and Olivia S. Mitchell, Strengthening Employment -Based Pensions in Japan, Benefits Quarterly, 2 nd Q, 2002: Cogan, John and Olivia S. Mitchell. Perspectives from the President s Commission on Social Security Reform. Journal of Economic Perspectives, Hsin, Ping-Lung and Olivia S. Mitchell. "Public Pension Plan Efficiency" In Positioning Pensions for the 21st Century. M. Gordon, O.S. Mitchell, and M. Twinney, eds. Pension Research Council. Philadelphia, PA: Univ. of Pennsylvania Press, 1997: James, Estelle, James Smalhout and Dimitri Vittas. Administrative Costs and the Organisation of Individual Account Systems: A Comparative Perspective. In New Ideas in Old Age Security: Towards Sustainable Pension Systems in the 21 st Century, R.Holzmann and J.Stiglitz: Mitchell, Olivia S. "Administrative Costs of Public and Private Pension Plans". In Privatizing Social Security, Martin Feldstein ed. NBER. Chicago: University of Chicago Press, 1998: Mitchell, Olivia S. and Emily Andrews. "Scale Economies in Private Multi-Employer Pension Systems." Industrial and Labor Relations Review 34 (July 1981): Mitchell, Olivia S. and Ping-Lung Hsin. "Managing Public Sector Pensions". In Public Policy Toward Pensions, J. Shoven and S. Schieber, eds. Twentieth Century Fund. Cambridge, MA: MIT Press, 1997: Mitchell, Olivia S. and Ping-Lung Hsin. "Public Sector Pension Governance and Performance". In The Economics of Pensions: Principles, Policies, and International Experience. Salvador Valdes Prieto, ed. Cambridge: Cambridge University Press, 1997: Mitchell, Olivia S. and Edwin Hustead. Pensions for the Public Sector. Pension Research Council. Philadelphia, PA: University of Pennsylvania Press, Murthi M, M.Orszag and P.Orszag, Administrative Costs Under a Decentralized Approach to Individual Accounts: Lessons from the United Kingdom, In New Ideas in Old Age Security: Towards Sustainable Pension Systems in the 21 st Century: Whitehouse, Edward. Administrative Charges for Funded Pensions: An International Comparison and Assessment. Pension Reform Primer Series, Social Protection Discussion Paper, No. 0016, The World Bank, Washington, 2000.

15 13 Table 1: International Comparison of Pension Plan Structures Country Features of Pension Industry Argentina Voluntary Defined contribution plans only Retail plans open to the general public Australia Mandatory and voluntary Trend towards defined contribution plans Plans open to the public (retail) or restricted to employees of a firm or industry (employersponsored) Single employer and multi-employer plans Chile Mandatory (some voluntary contributions) Defined contribution plans only Retail plans open to the general public Japan Voluntary Partially funded company defined benefit plans (Employee Pension Plans and Tax Qualified Pension Plans) and unfunded severance payments Pension Reform Bill 2001 reforms encourage company defined contribution plans, reform company defined benefit plans and introduce personal defined contribution plans Mexico Mandatory Defined contribution plans only Retail plans open to the general public Singapore Mandatory One single public sector fund Defined contribution plans only Sweden Mandatory contributions to the Premium Pension System (defined contribution) and quasimandatory occupational pensions (defined benefit) Premium Pension System (retail plans), occupational plans (employer-sponsored) Switzerland Mandatory and voluntary Defined benefit and defined contribution plans Mainly employer-sponsored plans United Kingdom Voluntary (with incentives to contract out ) Company plans and personal pensions Defined benefit, defined contribution and money purchase United States Voluntary Company plans (mainly single employersponsored in the public and private sector) Trend towards defined contribution and money purchase plans Source: Bateman et al. (2001), Clark and Mitchell (2002).

16 14 Table 2: Characteristics of Australian Pension Plans Industrial Structure Plan Type Sponsor Type Open Closed DB DC Employer-based plan (N=1674) 9% 91% 49% 51% Single-employer (corporate) (N=1520) Multi-employer (industry) (N=116) Public sector employer (N=38) Retail plan (N=131) Other (N=115) Total (N=1920) Source: Authors tabulations from APRA plan data; see text. Dollar values are $A, where $1A=$US0.60 in Table 3: Australian Pension Plan Annual Administrative Expenses Plan Characteristic Plan type DB DC Sponsor type Retail $A/Member $A/Assets (%) Mean Median Mean Median $ $ % % 0.93 Employer(s) Average Source: Authors tabulations from APRA plan data; see text. Dollar values are $A, where $1A=$US0.60 in 1999.

17 15 Table 4. Regression Estimates of Administrative Expenses on Pension Plan Characteristics (Standard errors in parentheses) Explanatory Variable Dep. Var = (ln) Admin. Expenses Model 1 Model 2 (a) Model 3 (b) Plan Scale Assets (ln $A) 0.458** 0.460** 0.434** (0.028) Participants (ln) 0.373** (0.023) Plan Type DB (vs DC) 0.335** (0.053) Sponsor Type Retail (vs employer -sponsor) 0.703** (0.107) Operational Variables Member contributions (% of total) ** (0.090) Transfers in (% of assets) ** (0.172) External account management (vs in-house) (0.044) Externally managed assets (%) (0.001) Member investment choice (vs no choice) 0.08 (0.056) Benefit payments (% of assets) 0.263** (0.080) Transfers out (% of assets) 0.464** (0.169) External review (vs no review) (0.094) (Continued next page) (0.036) 0.374** (0.028) (0.751) 0.716** (0.117) ** (0.148) (0.203) (0.058) 0.001** (0.001) (0.067) 0.220** (0.084) 0.401** (0.172) (0.124) (0.029) 0.398** (0.024) 0.339** (0.053) ** (1.188) (0.095) (0.214) (0.045) (0.001) (0.057) 0.696** (0.177) 0.724** (0.172) (0.097)

18 16 (Continued from previous page) Explanatory Variable Interaction variables Dep. Var = (ln) Admin. Expenses Model 1 Model 2 (a) Model 3 (b) Plan Scale Assets (ln $A) x (DB or retail) 0.06 (0.060) Participants (ln) x (DB or retail) (0.052) Plan Type DB (vs DC) x (retail) 0.436** (0.091) ** (0.086) (0.966) Sponsor Type Retail (vs employer -sponsor) x (DB) (0.972) Operational Variables Member contributions (% of total ) x (DB or retail) 0.623** (0.186) ** (0.281) Transfers In (% of assets) x (DB or retail) (0.389) (0.389) External account management (vs in-house) x (DB or retail) (0.089) (0.181) Externally managed assets (%) x (DB or retail) (0.001) (0.002) Member investment choice (vs no choice) x (DB or retail) (0.122) 0.815** (0.238) Benefit payments (% of assets) x (DB or retail) (0.306) ** (0.203) Transfers Out (% of assets) x (DB or retail) 0.0** (0.0) 0.0** (0.0) External review (vs no review) x (DB or retail) (0.190) (0.360) Number of plans R Notes: (a) All Model 1 variables interacted with DB plan type. (b) All Model 1 variables interacted with retail sponsor type. **Coefficient statistically significant at the 5% level.

19 17 Table 5: Predicted Pension Administrative Expenses by Plan Size and Design Plan Type Median Small Large $A/yr % assets $A/yr % assets $A/yr % assets Employersponsored plan DC DB Retail Notes: Values obtained from Table 4 (Model 1) coefficients, holding all else constant other than plan size, type, and sponsor type. The median plan has 238 members and $A9.7 M in assets; small plans have 50 members and $A2 M assets; and large plans have 50,000 members and $A2 B in assets. See also Table 3.

20 18 Figure 1: Pension Plan Administrative Expenses per Member/yr ($A) Frequency Average = $A , ,000 Plan administrative expenses per member/yr ($A) Figure 2: Pension Plan Administrative Expenses as % of Assets Average = 1.08% Frequency Annual administrative expenses (% assets)