Superannuation fees,asset allocation and fund performance

Abstract

Superannuation fees have come under public scrutiny in recent years with the belief they are too high. We examine the determinants of fees and their relationship with fund performance. Superannuation funds with higher investment fees have higher allocations to asset classes that trade in more complex markets. For-profit retail funds charge both higher investment and administration fees than other funds. Funds with higher investment fees do not generate higher returns than the least expensive funds. These findings suggest that superannuation members may not earn higher after-fee risk-adjusted returns by holding funds with higher fees.

JEL Classification:

G11

Keywords

Administration fees asset allocation investment fees superannuation funds

1. Introduction

Australian compulsory superannuation industry is estimated at $A3.4 trillion as at March 2022,¹ making it an imperative source of retirement income for Australians. Superannuation fund fees are potentially a major contributor to the erosion of superannuation savings over the working life of an individual (e.g. Australian Government, 2020; Commonwealth of Australia, 2010a, 2010b), hence, it is important to recognise how these fees will affect the net wealth of members, and whether members are adequately rewarded for incurring higher fees. The importance of superannuation fees has been a focal point of the 2014 Australia Financial Systems Inquiry’s review of the superannuation system (Commonwealth of Australia, 2014), with a consensus emerging that the fees members are paying for fund products are too high. Minifie (2014b) argues that superannuation fees in Australia are three times the median Organisation for Economic Co-operation and Development (OECD) rate, on average. The Financial System Inquiry (Commonwealth of Australia, 2014) also argues for increased efficiency in superannuation. In other words, fees should be reduced. The Productivity Commission (2018) investigated superannuation efficiency and competitiveness and echo these sentiments, arguing that fees are excessively eroding member balances.

Whether these fees are excessive depends on the value that one derives from the superannuation product being purchased. Although we are unable to directly observe the non-performance value² that an individual superannuation fund member derives from the fees paid, we can link the investment and administration fees to observable fund characteristics to ascertain what factors can explain differences in fees among funds. Specifically, we examine how fund size, asset allocation, risk category and fund type influence investment and administration fees. Previous research has not precisely linked the investment fee to the asset allocation of each superannuation fund in Australia. Minifie (2014a) discusses investment fees charged by superannuation funds and concludes that they are too high. We believe the detailed data used in this study will shed additional light on whether investment fees are indeed excessive. We are also able to determine whether these fees have any relation with the risk-adjusted returns that superannuation funds are able to generate.

Using a comprehensive Australian superannuation fund data set that spans the period from June 2007 to March 2015, we first sort funds into five groups based on their investment fee and show that investment fees are nearly three times higher for the more expensive funds compared to the least expensive funds. A similar exercise based on administration fees shows that administration fees are more than five times higher for the top 20% of funds when grouped by administration fee compared to the bottom 20%.

Our results reveal that exposure to certain asset classes helps to explain the investment fee. We show that this difference provides an intuitive reason as to why investment fees differ between funds. Funds with exposure to riskier asset classes that are less liquid and operate in complex markets, such as private equity and hedge funds, charge higher investment fees. A second contribution of the study is that fund type plays an important role in the determination of fees. After taking into account the asset allocation for each fund, we find convincing evidence that there are differences in the fees charged by corporate funds, industry funds, retail funds and public funds. These fee differences exist across both investment and administration fees. We find that industry funds have investment fees that are over 12 basis points cheaper than retail funds, after controlling for asset allocation. This is a significant amount, given that the average investment fee in our sample is around 60 basis points. Part of this 12 basis points difference is likely to reflect the fact that retail funds are for-profit and therefore need to build in a profit margin. The remainder of the funds in our sample are not-for-profit funds that do not need to build in such a profit margin. An additional issue to note is that we are examining headline fees. It is possible that larger companies may be able to negotiate discounts on behalf of their members. This behaviour is likely to be more prevalent in the for-profit sector.

In terms of the link between fees and performance, we do not find evidence that high-fee funds generate high returns. Neither, do we find evidence that high-fee funds generate lower returns. Part of the explanation for these higher fees is that these funds hold asset classes that trade in complex and illiquid markets, such as hedge funds, equity and infrastructure, which cost more to manage.³ We find that the difference in after-fee abnormal performance is not significantly different between the group of funds with the highest and lowest fees. This suggests that superannuation members were no better (or worse) off than they would have been if they had selected either the high- or low-fee fund during our sample period. In a similar result to what we observe for the determinants of fund fees, we find that retail funds also exhibit underperformance of risk-adjusted returns, when compared to public sector and corporate funds.

Our results are consistent with the work of Coleman et al. (2006) for Australian superannuation funds, with regards to fund type being the primary determinant of investment performance.⁴ Our results contrast with Basu and Andrews (2014) who find a negative relationship between fees and performance for funds on aggregate, a positive relationship between fees and performance for industry funds, and for retail funds the relationship is negative. Drew and Stanford (2003) argue that the current structure of superannuation funds leads to a series of agency-related problems. Drew and Stanford (2003) also summarise the then-prior empirical literature on the performance–fee relationship that shows that abnormal returns of funds are (generally) negatively related to the level of fees they charge. Rice Warner Actuaries (2014) show that assets under management grew by 11.1% over the period June 2004–June 2013, whereas, fees fell by only 0.20% from 1.40% to 1.20% over this same period. Chant West (2014) argues that the 10 largest non-profit active funds’ net returns significantly outperformed the passive benchmark by 0.4% per annum during the 10 years to December 2014, and that higher fees and higher returns can go together. They subsequently argue that expensive funds justify the higher fees that are charged to members by having higher returns. Chant West (2014) differs to our analysis as they use the passive Vanguard growth fund as the benchmark, rather than a benchmark based on the strategic asset allocations of funds. The Productivity Commission (2018) shows that retail funds underperform not-for-profit funds, but do not statistically attempt to link performance and fees.

Our article represents a more comprehensive study of fees and performance than recent Australian studies on superannuation. First, we use monthly returns’ data that are at a higher frequency than most studies. For example, Coleman et al. (2006) and Cummings (2016) use quarterly returns, while Basu and Andrews (2014) use annual returns. Granularity of returns is important, particularly when testing for trading performance when information is short-lived (e.g. Chen et al., 2017; Kothari and Warner, 2001).⁵ We also use the strategic asset allocations of superannuation funds and divide the allocations into 12 distinct asset classes. This allows us to benchmark funds more accurately based on their target or strategic asset allocations, which in turn may change over time. Previously, only Basu and Andrews (2014) benchmark funds based on their strategic asset allocations but use only six asset classes. Our use of 12 asset classes allows us to more properly benchmark funds especially in alternative assets where we obtain allocations separately for hedge funds, infrastructure and private equity. Prior studies, such as Coleman et al. (2006) and Cummings (2016), do not have information on the asset allocations of superannuation funds. Instead, they attempt to decompose allocations through a regression of fund returns using less than seven asset class benchmarks. Such a decomposition is subject to error to the extent that the set of benchmark indices is unrepresentative of the opportunity set faced by the funds, and hence may not accurately reflect the actual or strategic allocations of assets that are contained with the funds. The Productivity Commission (2018) presents charts on fees, performance and asset allocation but do not rigorously attempt to link these three areas.

The findings from this article are important for superannuation members and advisors when making informed investment decisions and when allocating funds between superannuation products. It is also relevant for superannuation fund managers when deciding on how to set fees in a competitive setting based on their asset allocations and relative peer performance. For example, funds with low risk allocations are expected to set lower fees reflecting the ease of managing such assets. Finally, the findings from this article are relevant for policy makers, who in recent years have promoted greater competition in the superannuation industry and placed pressure on funds to reduce fees (see Australian Government, 2020). Our recommendations from these findings are that superannuation members should be cautious when making investment decisions based solely on fee levels because fees cannot be used to determine the ability of fund managers. Furthermore, fund managers should not be forced to reduce fees to remain competitive, as this may lead to divestment from certain asset classes which provide higher returns, thus adversely affecting retirement balances.

There are two caveats to this study. First, we do not have data on the universe of superannuation funds. We believe that the availability of asset allocation data far outweighs the cost of only examining a smaller sample of superannuation funds. Second, the heterogeneity of investors makes it difficult to examine qualitative factors that will impact on the level of administration fees. The provision of other services, such as financial advice, may not be in the best interests of all members as many will not utilise these services. Therefore, the fee charged by a fund may be reasonable for one member, but not for a different member. The approach we take is to exclude qualitative factors (e.g. financial planning advice, rollover and consolidation services and retirement planning advice) from our analysis, given the value derived is subjective.

2. Hypotheses

This study focuses on the asset allocation of each fund, risk category, the size of its assets under management and the type of fund to identify the factors that impact the investment and administration fee charged by superannuation funds in Australia. We then link these fees to performance to ascertain whether there is any benefit to superannuation fund members who invest in high-fee funds.

A fund’s asset allocation is widely considered to be an indicator of the fees it charges. Specifically, the operating costs of a fund are likely to be influenced by the classes of assets held by the fund. Haslem et al. (2008) show that fees differ across Morningstar style classifications,⁶ whereas Adams et al. (2012) and Nanda et al. (2009)⁷ argue that multi-asset class funds are more expensive than funds that focus on a single asset class. Dellva and Olson (1998) show that funds with an international focus tend to be significantly more expensive than domestic-focused funds, that there are economies of scale with respect to fees and that high turnover funds have higher fees, whereas Geranio and Zanotti (2005) argue that among the foreign domiciled fund operating in Italy, fund-of-funds are the most expensive, followed by international equity funds, then balanced, high-yield bond, bond and cash funds.⁸ Khorana et al. (2009) in a comprehensive study of 46,850 mutual fund from 18 countries using fee data for 2002 find that ‘fees are lower for larger funds and fund families, index funds, funds of funds, guaranteed funds, and funds that require a higher minimum investment’ (p. 1307). Wahal and Wang (2011) argue that increased competition in markets through the entry of substitute funds is successful in lowering fund management fees. They argue that the US market for equity funds is competitive. Cooper et al. (2011) contest this finding, showing that entry into the market by similar funds does not impact upon expenses. Funds that are managed more actively are likely to incur higher operating costs, and as such, are expected to pass these costs on to investors by charging higher fees. Alternatively, funds that more closely track the market offer less scope to achieve returns that greatly exceed this benchmark. We expect fund fees to be positively related to the level of active management. Iannotta and Navone (2012) investigate this relationship and show that the US funds with a higher active share, measured by their four-factor R², charge higher fees. In addition, funds that are of higher risk and funds that invest more in small capitalisation stocks charge higher fees. Basu and Andrews (2014), however, show that Australian retail superannuation fund fees are negatively related to their level of active management, whereas the relationship is positive for industry funds.

We hypothesise that funds with a higher asset allocation to asset classes that trade in more complex markets (i.e. hedge funds, infrastructure and private equity) will have a higher investment fee. We do not have any prior conjecture on how asset allocation will impact the administration fee. We also investigate how fees vary within different risk categories as an alternative way of proxying for funds exposure to more expensive asset classes. We anticipate that funds in the higher risk categories, and with greater exposure to asset classes that trade in complex markets, will have higher investment fees given the additional cost associated with managing these less liquid assets. We expect investment fees to increase across fund type from conservative, balanced, growth and high growth.

A commonly investigated factor sought to explain fees is the size of a fund’s assets under management. A number of studies show that a significantly negative relationship exists between the size of US mutual funds and fees charged to investors (Berkowitz and Kotowitz, 2002; Christoffersen, 2001; Dellva and Olson, 1998; Gil-Bazo and Ruiz-Verdú, 2009; Golec, 2003; Luo, 2002;⁹ Malhotra and McLeod, 1997; Tufano and Sevick, 1997). Khorana et al. (2009) observe a negative relationship between size and expenses for an international sample of managed funds. Martínez Sedano and Gil-Bazo (2004) study Spanish mutual funds and find that fees are significantly negatively related to fund size, when fund size is proxied by assets under management per fund investor. Geranio and Zanotti (2005) examine Italian mutual funds and find that fees are significantly negatively related to total net assets and total net assets managed by the same asset management company. The inverse size–fee relationship is consistent with the notion of economies of scale, in that, smaller funds are required to increase fees to cover fixed costs that would have otherwise been spread across larger funds under management. In rare contrast to these findings, Lesseig et al. (2002)¹⁰ find that size is positively related to the management fees of managed funds in the United States. Martínez Sedano and Gil-Bazo (2004) also show that fees for all objective groups of non-guaranteed funds are significantly higher than funds that have a short-term fixed-income objective, demonstrating that the fee structure differs over the fund’s mandate.

The literature on the relationship between fund size and fund performance is mixed. Bauer et al. (2010) find that smaller funds outperform their benchmark, while Andonov et al. (2012) show that there are substantial diseconomies of scale that are directly related to illiquidity. Dyck and Pomorski (2011) show that there are positive-scale economies between size and returns. Evidence from studies of Dutch pension plans is also inconclusive. Bikker (2017) finds no diseconomies of scale for larger funds, while Alserda et al. (2018) find large economies of scale for administrative expenses but modest diseconomies of scale for investment activities. Phillips et al. (2018) argue that the literature on economies of scale is ambiguous because endogeneity has not been recognised in the prior research. They use a set of instrumental variables to correct for the endogenous relationship between fund size and fund performance, and conclude that size does not affect performance.

The previous results on economies of scale in the fees charged by Australian superannuation funds are also mixed. Bateman and Mitchell (2004), Sy (2010),¹¹ Higgs and Worthington (2012), Basu and Andrews (2014) and Cummings (2016) show that scale economies exist among Australian superannuation funds. Liu and Arnold (2010) show that larger funds benefit from economies of scale, especially when their services are outsourced to independent providers. On the other hand, Coleman et al. (2006) find minimal evidence of a relationship between fund size and fees. Tan and Cam (2014) investigate fee structures within various governance structures for Australian not-for-profit pension funds. Their findings suggest that the conventional good governance practices do not benefit superannuation fund members in terms of cost reduction. Despite this contradictory evidence in the work of Tan and Cam (2014) in relation to scale economies, we expect that the investment fee and the administration fee should decline as assets under management increase. The ability of superannuation funds to negotiate lower investment fees for large mandates, and the spreading of operational costs across a large amount of assets and members, should lead to lower fees as fund size increases.

There are a number of reports and studies that show differences in the level of fees across retail, industry, corporate and public sector superannuation funds (Coleman et al., 2006; Ellis et al., 2008; Minifie, 2014b). It is important to control for these factors. Previous studies have shown that retail funds have relatively high fees and corporate and public sector funds have relatively low fees. Our research will investigate whether these fee differences exist after controlling for the asset allocation of funds.

The relationship between fund fees and risk-adjusted performance is mixed, at best. Coleman et al. (2006) find no relationship between fees and performance for superannuation funds when performance is measured using the alpha from a multi-index model, and a significant negative association when performance is measured using either return on assets or the information ratio. The US evidence from the work of Tufano and Sevick (1997), Spanish evidence from the works of Martínez Sedano and Gil-Bazo (2004), Italian evidence from the work of Geranio and Zanotti (2005) and Finnish evidence from the work of Korkeamaki and Smythe (2004)¹² indicate that fees do not impact performance. There is, however, an abundance of literature that conflicts with these findings.

In the United States, Lesseig et al. (2002)¹³ and Berkowitz and Kotowitz (2002) show that funds with higher expense ratios outperform the less-expensive funds. Basu and Andrews (2014) observe a positive relationship between fees and performance for retail funds, a negative relationship for industry funds and no association for corporate funds when examining Australian superannuation funds. The US studies on mutual fund performance largely show an inverse relationship with respect to fund expenses (see, for example, Carhart, 1997; Cooper et al., 2011; Dellva and Olson, 1998; Gil-Bazo and Ruiz-Verdú, 2009; Gruber, 1996; Haslem et al., 2008). Bechmann and Rangvid (2007) similarly show, in the Danish market, that the cheapest fee funds outperform the most expensive fee funds across long hold-out periods,¹⁴ while Otten and Bams (2002) show that fees are negatively related to performance in three (the United Kingdom, Germany and Netherlands) of the four (no significant relationship is found for France) European nations they examine.¹⁵ Mehri et al. (2021) compare the fees charged by Islamic and conventional funds and conclude that legal, cultural and political factors help explain the differences in fees.

Iannotta and Navone (2012) and Christoffersen and Musto (2002) suggest that underperforming mutual funds are able to set higher fees due to the performance in-sensitivity of the clientele they attract. In other words, lower performing funds are able to charge higher fees without the risk of incurring asset outflows. However, there is also evidence that defined contribution fund investors are more performance-sensitive than defined benefit plan investors (e.g. Sialm et al., 2015). Gil-Bazo and Ruiz-Verdú (2009) similarly argue that underperforming mutual funds cannot compete for the same investors as the best performing funds, so that they market themselves to the less sophisticated, less performance-sensitive investors and are thus able to charge higher fees.¹⁶ Given this mixed evidence, we do not have a firm prediction of whether fund fees will impact fund performance.

3. Data

The superannuation fund data used in this study are sourced from fund surveys collected by Chant West.¹⁷ Chant West is an independent superannuation research and consultancy firm established in 1997. The sample covers annual data for the period from 2007 to 2015. The data extracted from these surveys include portfolio returns after investment fees, percentage investment fees and the dollar value of administration fees for account balances of $25,000, $50,000 and $250,000. The investment fee covers the cost of managing the investments of the superannuation fund. The administration fee covers the costs incurred to operate the superannuation fund. The dollar amount of the administration fee charged varies across the three different balance amounts. The percentage administration fee also varies by account balance. The total fee is the investment fee plus the administration fee. We convert the dollar value of the administration fee to a percent value by dividing by the account balance. For the majority of the analysis, we focus on an account balance of $50,000 as the results are not substantively different for the $25,000 and $250,000 account balances. The data set also includes strategic asset allocation across Australian shares, Australian bonds, international shares, international bonds, Australian property, international property, infrastructure, hedge funds, cash, private equity and other assets. Size is measured at the superannuation fund level and not individually for each investment option within a fund. Information on fund provider type (retail, corporate, industry and public-sector funds) and an indicator for MySuper funds are also contained in the final data set.¹⁸

The sample used in the analysis requires funds to have observations for portfolio returns, investment fee, administration fee, asset allocation and the total assets under management of the superannuation fund (analysis sample). These constraints produce an unbalanced panel of 992 annual observations across 146 superannuation fund options. This represents about half of the funds in the Chant West surveys. In untabulated tests, we compare the fees and returns of superannuation funds that do not satisfy our sample selection criteria with those that do (where fees’ and returns’ data are available). The differences between these two groups are minimal and we conclude that our sample is representative of Chant West’s data universe. Furthermore, we compare our filtered sample to the Australian Prudential Regulation Authority’s (APRA) annual fund-level superannuation statistics in panel A of Appendix 1 and find generally good coverage. We do note that there are differences in fund size between the Chant West sample we analyse and the APRA sample. The assets under management of the funds in our sample are considerably larger than the APRA sample (see Appendix 1, panel B). As a result, our results may not generalise to smaller superannuation funds.

Table 1 contains descriptive statistics on the fee variables that are observed at an annual frequency. The average investment fee is just under 62 basis points in our sample. The median is similar, though there is significant variation in the fee within the sample. In terms of the administration fee, these average from 62 basis points on account balances of $25,000 to an average of 16 basis points for account balances of $150,000. The medians are lower for all the administration fees, suggesting that there are some funds charging sizable administration fees. The maximum values confirm this. Investment fees make up 50% of the total fee for account balances of $25,000 and 79% for account balances of $150,000. Table 2 presents the mean and median fee for each year from 2007 to 2015. There is considerable variation in investment fees over time, with an increasing trend evident up until 2010 followed by some volatility. The median follows a similar pattern. The times series behaviour of the administration fee is essentially the opposite. The mean administration fee declines over the sample, before increasing from 2013 to 2015. The median results are broadly similar. We will control for this time series variation using time-fixed effects in the regression analysis.

Table 1.

Fee descriptive statistics. Descriptive statistics are presented for a sample taken from the Chant West data set for 992 funds with complete data on fees, fund size and asset allocation.

Investment fee (%)		Administration fee (%)				Total fee (%)
		$25,000	$50,000	100,000	$150,000	$25,000	$50,000	$100,000	$150,000
Mean	0.615	0.615	0.476	0.238	0.159	1.229	1.090	0.853	0.773
Median	0.620	0.500	0.360	0.180	0.120	1.150	1.030	0.840	0.768
Std. dev.	0.227	0.360	0.332	0.166	0.111	0.434	0.410	0.287	0.257
Min.	0.110	0.210	0.100	0.050	0.033	0.380	0.260	0.200	0.180
Max.	2.630	4.850	2.450	1.225	0.817	5.510	3.640	3.135	2.967

Table 2.

Fee descriptive statistics by year.

	Investment fee (%)				Administration fee (%)				Total fee (%)
	Mean	Median	Min.	Max.	Mean	Median	Min.	Max.	Mean	Median	Min.	Max.
2007	0.582	0.610	0.170	1.020	0.508	0.560	0.100	1.300	1.090	1.140	0.350	1.740
2008	0.594	0.610	0.110	1.490	0.491	0.390	0.100	2.450	1.085	1.000	0.290	3.110
2009	0.622	0.610	0.110	2.630	0.458	0.300	0.100	1.300	1.080	1.030	0.260	3.640
2010	0.653	0.615	0.190	2.630	0.433	0.280	0.100	1.300	1.086	1.035	0.310	3.640
2011	0.653	0.660	0.170	1.290	0.426	0.280	0.100	1.300	1.078	1.060	0.340	1.860
2012	0.622	0.650	0.170	1.190	0.451	0.342	0.100	1.300	1.073	1.040	0.310	1.940
2013	0.590	0.600	0.180	1.100	0.479	0.360	0.140	1.470	1.068	0.995	0.480	2.090
2014	0.605	0.610	0.249	1.040	0.519	0.410	0.130	1.790	1.124	1.015	0.520	2.750
2015	0.598	0.600	0.249	1.040	0.522	0.430	0.130	1.790	1.119	1.010	0.590	2.750

The mean, median, minimum and maximum of the investment, administration and total fee are reported on an annual basis for funds from the Chant West data set with complete data on fees, fund size and asset allocation. The administration fee and total fee are based on an account balance of $50,000.

The benchmark indices that we use are in Appendix 2. Aside from the private equity and hedge fund indices, these indices are investible through a position in passive funds that benchmark against such indices. We were unable to obtain unlisted property and unlisted infrastructure fund benchmarks as these benchmarks are either not useful for benchmarking or difficult to obtain. For unlisted property, IPD MSCI provides date on unlisted Australian and the US property indices. However, in the Bloomberg description of the indices, it states that these indices are ‘neither appropriate nor authorized for use as a benchmark for portfolio or manager performance’. As such, we use the equivalent real estate investment trust (REIT) indices for unlisted property allocations. For infrastructure, the closest measure of unlisted infrastructure that we could find is in the work of Bird et al. (2014) who use an equal-weighted net-of-fees return data series for 10 unidentified unlisted infrastructure managers. As these data are not publicly available and may be subject to selection bias, we choose to use infrastructure REIT indices as benchmarks for unlisted infrastructure. Benchmark accumulation index returns for the major asset classes invested in by the superannuation funds were retrieved from Thomson Reuters Tick History and Bloomberg. The return of the riskless asset is the cash rate from the Reserve Bank of Australia (RBA) website.

4. Method

This study uses two different approaches to investigate the factors that influence the level of fees charged by superannuation funds and the link between fees and performance. The first approach forms portfolios ranked on the level of fees to determine whether there are differences in characteristics and performance across funds with different fee levels. The second approach uses regression analysis to identify what factors are responsible for both investment fee and administration fee levels. Separate regressions examine the link between fees and performance.

4.1. Fee determinants: portfolio sorts

The portfolio approach sorts superannuation funds into five groups each year based on their investment fee. We calculate the mean and median of each portfolio’s asset allocation to identify if differences in fees are related to asset allocation. The same analysis is conducted for portfolios sorted on administration fees. We also undertake a separate analysis where we form three investment fee-sorted portfolios using sub-samples based on the investment option (i.e. conservative, balanced, growth and high growth). This analysis is designed to identify if differences in fees vary across investment options, and if asset allocation also has a role to play within each investment option. We calculate the difference in the asset allocation proportion between the high- and low-fee portfolios to gain a preliminary understanding of the relationship between fees and asset allocation. One benefit of the portfolio approach is that we can identify non-linear relationships.

4.2. Fee determinants: regression analysis

To gain a better understanding of the factors that help explain the level of fees set by superannuation funds, we estimate various regressions using fund-level panel data measured at an annual horizon. We employ investment and administration fees as dependant variables in separate regressions. All fees are measured in percent, with the administration fee based on an account balance of $50,000. The independent variables are the percentage allocation for each fund in the following asset classes: Australian shares, Australian bonds, international shares, international bonds, Australian property, international property, infrastructure, hedge funds, cash and private equity assets. As a robustness test, we exclude hedge funds, infrastructure and private equity from the regression, and include the aggregate percent of the portfolio invested in alternative asset classes.

For robustness, we include a series of dummy variables that indicate the risk category to which each fund belongs and substitute these for the asset allocation variables. Chant West classifies funds as being conservative if they hold less than 40% of their allocation in growth assets, balanced if growth assets are between 40% and 60%, growth as 60% to 80%, and high growth funds have over 80% of their allocation in growth assets. Growth assets include equities, property, and certain alternative assets, such as hedge funds, private equity and infrastructure. We include dummy variables for conservative, growth and high growth funds, with the reference group being balanced funds. We include dummy variables for the different types of superannuation fund. Specifically, the variable Retail is equal to one if the fund is a master trust or a consultant and zero otherwise. We include similar variables for public-sector funds and corporate funds, with industry funds being the reference group. These dummy variables will estimate the fee difference relative to industry funds. Size is measured by the natural log of a fund’s net assets under management. We include a dummy variable that takes a value of one for a fund classified as a MySuper fund, or zero otherwise. Year-fixed effects are included in all the regressions and the standard errors are clustered at the fund level.

4.3. Benchmark adjusted fund returns

When adjusting superannuation fund returns for benchmarks, we rely on the strategic asset allocations of funds to create an asset allocation benchmark (AAB) for each fund. For each eligible fund, we calculate an AAB as the sum of lagged strategic asset allocation multiplied by the respective asset-class benchmark return. The AAB has several distinct advantages to other forms of benchmarking. First, it is dynamic as it follows the changing asset allocation of a fund. The benchmark is therefore customised to each fund at each point in time. In contrast, regressing returns on several benchmarks (such as in a factor model) does not capture funds changing allocations over time. Second, because the AAB uses lagged strategic asset allocations as the benchmark, it simultaneously captures performance from increased returns through deviating from its strategic asset allocations (i.e. undertaking tactical asset allocation), as well as capturing excess returns within individual asset classes (security selection). Gallagher (2001) uses a similar approach.

4.4. Performance determinants: portfolio sorts

To investigate the association between superannuation fund performance and their level of investment fees, we construct fee-sorted quintile portfolios from all superannuation funds with monthly returns, investment fees and 1-month lagged strategic asset allocations. For each portfolio, monthly equal-weighted returns are calculated over the period from June 2007 to March 2015. The portfolio is rebalanced monthly. Using this method, we simulate the average actual performance of unit holders/fund members in high- or low-fee funds at a given point in time. As such, there is no hindsight bias in our estimation. Also, as fund fee-group returns are compared at the same points in time, there is no need to correct for time trends in fees. For example, fees could be falling over time due to lower fees charged by investment managers, administration or research costs; this should be observed in all groups. If we, however, find that some funds do not pass on the cost savings in the form of lower fees, then we should expect to find a negative relation between fund performance and after-fee returns.

4.5. Individual fund regression

Using fund fee-group portfolios ignores potential fund characteristics, such as fund type that may also affect fund performance in addition to fund fees. As such, individual fund benchmark-adjusted returns are regressed against investment fees and other fund characteristics to identify the relationship between superannuation fund performance and their investment fees. The full regression model that is estimated is

α_{i, t} = β_{0} + β_{1} {Fee}_{i, t} + \sum_{g = 1}^{3} γ_{g, i} {Group}_{i, g} + β_{3} {Size}_{i} + β_{4} {MySuper}_{i} + F_{i} + T_{i} + ε_{i, t}

(1)

where α_i,t is the monthly performance of fund i at time t. We use the fund return in excess of the risk-free rate and the fund return above its AAB (abnormal returns) to measure performance. Fee is either the fund-level investment fee in percentage per annum or the administration fee on a $50,000 account balance converted to percentage per annum. Size measures the assets under management for the particular fund, MySuper is a dummy that takes the value of one if the fund is a MySuper fund, or zero otherwise. Group, is a vector of three indicator variables that take the value of one if the fund category type is either one of retail, corporate fund or industry fund, and zero otherwise. The omitted category is industry funds. Fi and Tt are fund- and time-fixed effects, respectively. ω is the intercept term and ε_i, is the error term. Standard errors are clustered by year/month.

5. Results

This section reports the findings from our empirical analysis. Sections 5.1 and 5.2 present results on the determinants of superannuation fees and attempts to answer the question as to whether fees are excessive. They also seek to determine the characteristics of funds that charge higher fees. Section 5.3 reports various fund performance measures for the different fee portfolios. The regression results on the determinants of fund performance are reported in Section 5.4.

5.1. Asset allocation differences between fee-sorted portfolios

Table 3 contains the portfolio results, with panel A presenting the investment fee-sorted portfolios and panel B containing the administration fee-sorted portfolios. In panel A, the average investment fee for the high-investment fee portfolio is almost 91 basis points, which is 57 basis points higher than the 34 basis point investment fee charged by the low-fee portfolio. This is an extremely large difference. However, there could be economic justifications for this difference. The administration fee on a $50,000 account balance is not significantly different for funds with higher investment fees. This provides some evidence that investment fee and administration fee are set independently. Not surprisingly, the total fee varies in a monotonic fashion from the low-fee portfolio to the high-fee portfolio. A large number of the asset allocations change in a monotonic fashion, as expected. Funds with higher investment fees have a lower allocation to cash, Australian bonds and international bonds. Funds with higher investment fees have a higher allocation to Australian shares, international shares, Australian property, hedge funds, private equity and infrastructure. With the exception of international property, the funds with exposure to asset classes that trade in more complex markets charge higher fees. For example, low-fee funds have a 19% allocation to cash whereas high-fee funds have a 4% allocation. This 15% difference is roughly the same as the difference in the weights allocated to Australian shares between the low- and high-fee portfolios, with high-fee funds having a larger allocation to shares. When grouping all alternative asset classes together, we see that the funds with highest investment fee have more than twice the allocation to alternative assets (7% vs 16%). It is worth pointing out that the weight in alternative assets increases monotonically from the low-investment fee portfolio to the high-investment fee portfolio. At first glance, it seems that fees are in line with expectations given the risk and costs of the various asset classes. However, different investment options have different asset allocations that could influence the investment fee charged. We will return to those differences shortly.

Table 3.

Asset allocations of fee-sorted portfolios.

Fee Group	Investment fee	Admin. fee, $50,000	Total fee, $50,000	Cash	Aust. shares	Aust. bonds	Intl. shares	Intl. bonds	Aust. property	Intl. property	Hedge funds	Private equity	Infra.	Alt.
Panel A: investment-fee-sorted portfolios
Low	0.334	0.462	0.796	18.690	17.972	10.818	17.119	9.302	6.358	1.291	1.149	0.696	1.841	7.045
2	0.488	0.442	0.929	12.808	21.451	8.228	19.366	8.601	6.353	1.656	2.894	1.257	2.904	10.225
3	0.610	0.483	1.093	8.401	26.700	7.307	22.555	8.220	6.216	2.124	3.494	2.178	3.489	12.389
4	0.709	0.537	1.245	6.830	29.787	5.120	25.613	5.840	6.415	2.250	3.982	2.249	3.861	13.859
High	0.905	0.456	1.361	4.144	32.697	2.983	28.215	3.393	7.533	1.556	4.584	2.941	4.807	15.973
High–low	0.571***	–0.006	0.565***	–14.546***	14.725***	–7.835***	11.096***	–5.909***	1.174***	0.265	3.435***	2.245***	2.966***	8.928***
t-stat	(32.22)	(–0.17)	(14.00)	(–14.95)	(17.91)	(–9.24)	(12.38)	(–8.70)	(2.91)	(1.30)	(9.43)	(10.49)	(8.92)	(14.40)
Panel B: administration-fee-sorted portfolios
Low	0.580	0.157	0.737	9.383	25.318	5.134	23.265	7.112	8.438	0.508	2.926	3.022	5.140	14.714
2	0.592	0.234	0.826	9.633	26.915	5.599	21.363	4.228	7.728	1.260	2.526	2.288	4.477	11.836
3	0.673	0.352	1.025	11.132	26.334	5.198	21.829	4.497	7.793	1.404	3.614	1.888	3.656	12.617
4	0.564	0.574	1.138	11.539	23.694	7.575	23.330	8.015	6.901	1.876	2.986	1.417	2.652	10.130
High	0.658	0.998	1.656	8.481	26.841	10.200	23.785	11.234	2.497	3.567	4.163	1.024	1.445	11.168
High–low	0.078***	0.841***	0.919***	–0.902	1.524	5.065***	0.520	4.122***	–5.940***	3.059***	1.238***	–1.998***	–3.695***	–3.546***
t-stat	(2.89)	(42.87)	(28.35)	(–0.89)	(1.42)	(6.61)	(0.49)	(4.84)	(–20.62)	(18.73)	(2.71)	(–7.67)	(–11.38)	(–4.91)

Superannuation funds are ranked into five portfolios each year based on their fee. The portfolios are formed based on two fees – investment and administration. The administration fee is based on an account balance of $50,000. The three mean fees (investment, administration and total) and the mean asset allocations for each portfolio of superannuation funds are based on the average for the entire sample. The difference between the average of the high- and low-fee groups is reported in the high–low row with the associated t-statistic contained in parentheses. *, ** and *** indicate statistical significance at 10%, 5%, and 1%, respectively.

Panel B presents the fees and asset allocations across portfolios sorted by administration fee charged to an investor with a $50,000 account balance. There is a sizable difference in administration fees, ranging from 0.157% in the low-fee portfolio to 1% in the high-fee portfolio. This indicates that there is scope to lower administration fees. High administration fee portfolios have slightly higher investment fees but the average fee across the portfolios is not monotonic. High administration fee portfolios also have higher total fees. When examining the asset allocations, it is evident that funds with higher exposure to the two bond asset classes, international property and hedge funds have higher administration fees. There is a negative relationship for Australian property, private equity and infrastructure. Although we had no expectation about how administration fees would vary with asset class, it is unexpected that we find some strong relationships. The regression analysis will take into account other factors that could influence administration fees to ascertain the strength of these portfolio-level relationships.

To identify whether it is purely risky assets, or particular asset classes, that explain the difference in investment fee, we separate funds into four groups based on their investment option: conservative, balanced, growth and high growth. Within each investment option, we sort funds into three groups based on their investment fee, with rebalancing undertaken every year. These average fees and asset allocations are reported in panels A–D of Table 4. Inspection of the table shows that the investment fee and the portfolio weights in asset classes that trade in more complex markets differ between investment options, as expected. What is apparent is that the difference in fees that exists in the entire sample is present within each investment option group. The difference in investment fee between the high- and low-investment fee portfolios ranges from 34 to 42 basis points per annum across the four risk categories. Again, these differences are sizable. There are some consistent patterns that emerge across each of the four panels. First, within each investment option, those funds that had higher fees all had significantly higher allocations to hedge funds. In all cases, the high-fee funds exposure to hedge funds was double that of low-fee funds. The allocation to infrastructure is not monotonic within each investment option, but the higher fee funds all have a higher allocation to infrastructure than the low-fee funds. The key message is that higher investment fees are charged by funds that have a higher allocation to less liquid, asset classes that trade in complex markets. However, this analysis ignores other influences that could impact fund fees. Accordingly, we now turn our attention to regression analysis to determine whether the exposure to the illiquid asset classes can explain the fee differences when other possible explanations are introduced.

Table 4.

Asset allocations of investment-fee-sorted portfolios by risk category.

Fee group	Investment fee	Admin. fee, $50,000	Total fee, $50,000	Cash	Aust. shares	Aust. bonds	Intl. shares	Intl. bonds	Aust. property	Intl. property	HF	PE	Infra.	Alt.
Panel A: conservative
Low	0.293	0.485	0.778	27.798	11.632	14.191	9.791	11.760	5.610	1.232	1.149	0.389	0.987	6.002
Mid	0.435	0.372	0.807	25.376	11.928	10.717	10.378	10.498	5.655	0.844	2.839	0.460	3.396	8.824
High	0.658	0.590	1.249	24.897	12.147	12.776	9.502	15.302	4.838	1.721	4.111	0.334	2.142	9.742
High–low	0.365***	0.105*	0.471***	–2.902**	0.515	–1.415	–0.289	3.541**	–0.772	0.489*	2.963***	–0.055	1.155***	3.740***
t-stat	(13.46)	(1.83)	(7.31)	(–2.09)	(1.35)	(–0.80)	(–0.84)	(2.29)	(–1.30)	(1.96)	(5.48)	(–0.48)	(4.06)	(3.86)
Panel B: balanced
Low	0.357	0.492	0.849	11.695	19.532	12.395	17.542	10.273	6.999	1.221	1.338	0.837	1.827	8.040
Mid	0.538	0.440	0.978	11.695	19.929	7.548	17.610	9.368	6.671	1.419	4.141	1.383	3.317	12.473
High	0.699	0.533	1.232	13.758	20.654	9.125	18.002	9.864	6.703	1.404	3.799	1.534	3.009	11.149
High–low	0.342***	0.041	0.383***	2.063	1.122	–3.269*	0.460	–0.410	–0.296	0.182	2.461***	0.697**	1.183**	3.110***
t-stat	(23.50)	(0.59)	(5.35)	(1.63)	(1.51)	(–1.93)	(0.75)	(–0.30)	(–0.42)	(0.60)	(4.35)	(2.20)	(2.26)	(2.65)
Panel C: growth
Low	0.476	0.466	0.942	5.015	28.318	6.645	25.500	6.297	7.096	2.296	2.220	1.935	2.708	10.344
Mid	0.661	0.495	1.157	4.130	29.408	6.467	25.105	7.828	6.359	2.348	3.710	2.543	4.291	14.356
High	0.849	0.374	1.223	4.286	29.996	3.357	25.019	4.162	8.835	1.248	4.592	3.450	5.882	18.184
High–low	0.373***	–0.092**	0.281***	–0.729	1.678***	–3.288***	–0.481	–2.134***	1.739***	–1.049***	2.373***	1.515***	3.174***	7.840***
t-stat	(27.29)	(–2.00)	(5.79)	(–1.45)	(3.02)	(–4.79)	(–0.84)	(–3.14)	(2.96)	(–3.31)	(4.43)	(4.45)	(6.38)	(9.28)
Panel D: high growth
Low	0.537	0.496	1.033	1.372	37.165	1.756	35.036	1.793	6.985	2.339	1.674	2.131	3.018	10.367
Mid	0.710	0.511	1.222	0.857	37.740	2.344	32.917	1.643	6.498	2.580	4.048	2.947	4.013	14.502
High	0.953	0.484	1.438	1.068	37.844	1.137	33.532	1.529	6.380	1.950	4.555	3.038	4.509	15.372
High–low	0.416***	–0.012	0.405***	–0.304	0.679	–0.619**	–1.504*	–0.264	–0.606	–0.389	2.882***	0.907**	1.491***	5.005***
t-stat	(13.30)	(–0.22)	(6.26)	(–1.40)	(0.81)	(–2.10)	(–1.71)	(–0.70)	(–0.94)	(–0.95)	(4.67)	(2.31)	(2.88)	(5.56)

Superannuation funds are separated into groups based on their risk category (conservative, balanced, growth and high growth), and then ranked into three portfolios each year based on their fee. The portfolios are formed based on investment fee. The three mean fees (investment, administration and total) and the mean asset allocations for each portfolio of superannuation funds are based on the average for the entire sample. The difference between the average of the high- and low-fee groups is reported in the high–low row with the associated t statistic contained in parentheses. *, ** and *** indicate statistical significance at 10%, 5%, and 1%, respectively.

5.2. What are the determinants of superannuation fees?

The regression results on the determinants of fees are presented in Tables 5 and 6 for investment and administration fees, respectively. Column 1 of Table 5 shows that corporate and public-sector funds have lower investment fees compared to other funds. Retail funds and industry funds all have similar investment fees before any consideration is given to asset allocation. Column 2 indicates that the investment fee varies with risk categories as expected. Column 3 combines the risk category, fund type variables, fund size and the MySuper dummy variable to explain 35% of the variation in investment fees. Column 4 includes only the asset class weights and has an adjusted R² of 32%. The focus of our analysis is in column 5, and a number of important conclusions can be gleaned from the regression results.

Table 5.

Determinants of investment fees.

	1	2	3	4	5	6
Retail	0.011 (0.31)		–0.003 (–0.09)		0.089*** (2.65)	0.061* (1.67)
Corporate	–0.217*** (–5.51)		–0.218*** (–6.82)		–0.147*** (–4.29)	–0.167*** (–4.89)
Public sector	–0.180*** (–4.16)		–0.170*** (–4.84)		–0.120*** (–4.59)	–0.126*** (–4.70)
Conservative		–0.072* (–1.95)	–0.078** (–2.39)
Growth		0.129*** (3.94)	0.112*** (3.96)
High growth		0.203*** (5.24)	0.191*** (6.08)
Shares				0.465*** (6.42)	0.307*** (4.33)	0.367*** (5.89)
Bonds				0.065 (0.99)	–0.159** (–2.26)	–0.127 (–1.63)
Property				–0.089 (–0.23)	0.624* (1.81)	0.590* (1.74)
Infrastructure				1.088*** (3.40)	0.811*** (2.61)
Hedge funds				1.785*** (5.67)	1.025*** (4.08)
Private equity				1.273* (1.89)	1.711*** (2.74)
Alternatives						0.809*** (5.32)
Size			–0.022* (–1.75)		–0.019* (–1.68)	–0.019 (–1.59)
MySuper			0.016 (1.19)		0.013 (0.86)	0.014 (0.87)
Adj. R²	0.125	0.230	0.350	0.323	0.407	0.393
Observations	992	992	992	992	992	992

The dependent variable is investment fee in percent. The independent variables included are separate indicator variables that are equal to one for a given fund type and zero otherwise. These variables are included for retail funds, corporate and public-sector funds. Industry funds are the reference fund type. Separate indicator variables are included that are equal to one for a given risk category type and zero otherwise. These variables are included for conservative, growth and high growth risk categories. Funds with a balanced option are the reference risk category. The remaining independent variables are fund-level asset allocations measured in decimals (i.e. 10% is 0.1), the natural log of net assets under management for the superannuation fund (Size) and an indicator variable equal to one is the fund is classified as a MySuper fund, or zero, otherwise (MySuper). Year-fixed effects are included in all regressions. All regressions are with standard errors clustered at the fund level. t statistics are reported in parentheses. *, ** and *** indicate statistical significance at 10%, 5%, and 1%, respectively.

Table 6.

Determinants of administration fees.

	1	2	3	4	5	6
Retail	0.593*** (14.79)		0.565*** (15.18)		0.557*** (12.56)	0.577*** (11.53)
Corporate	0.060 (1.58)		0.059 (1.47)		0.055 (1.17)	0.058 (1.02)
Public sector	0.012 (0.30)		0.015 (0.38)		–0.006 (–0.13)	0.018 (0.42)
Conservative		–0.006 (–0.07)	–0.015 (–0.40)
Growth		–0.039 (–0.52)	–0.056 (–1.60)
High growth		0.014 (0.18)	–0.027 (–0.78)
Shares				0.595*** (5.84)	–0.048 (–0.70)	–0.061 (–0.90)
Bonds				0.763*** (5.26)	–0.008 (–0.11)	0.015 (0.18)
Property				–2.324*** (–3.98)	0.542 (1.31)	0.351 (0.80)
Infrastructure				–2.350*** (–4.58)	–1.586*** (–4.49)
Hedge funds				1.750*** (3.43)	0.538 (1.62)
Private equity				–1.890** (–2.28)	0.518 (1.14)
Alternatives						–0.019 (–0.06)
Size			–0.063*** (–5.66)		–0.045*** (–4.05)	–0.061*** (–5.54)
MySuper			0.035 (1.29)		0.017 (0.80)	0.008 (0.32)
Adj. R²	0.657	0.004	0.691	0.355	0.712	0.689
Observations	992	992	992	992	992	992

The dependent variable is the administration fee in percent based on an account balance of $50,000. The independent variables included are separate indicator variables that are equal to one for a given fund type and zero otherwise. These variables are included for retail, corporate and public-sector funds. Industry funds are the reference fund type. Separate indicator variables are included that are equal to one for a given risk category type and zero otherwise. These variables are included for conservative, growth and high growth risk categories. Funds with a balanced option are the reference risk category. The remaining independent variables are fund-level asset allocations measured in decimals (i.e. 10% is 0.1), the natural log of net assets under management for the superannuation fund (size) and an indicator variable equal to one is the fund is classified as a MySuper fund, or zero, otherwise (MySuper). Year-fixed effects are included in all regressions. All regressions are with standard errors clustered at the fund level. T-statistics are reported in parentheses. *, ** and *** indicate statistical significance at 10%, 5%, and 1%, respectively.

First, we observe that fund asset allocations significantly influence the level of investment fees that is charged. Specifically, the allocation to shares, property, infrastructure, hedge funds and private equity are significantly positively related to investment fees. The more complex asset classes of infrastructure, private equity and hedge funds have a much larger influence on investment fees. This result is intuitive given that these asset classes are typically considered high growth asset classes and are inherently riskier, trade in more complex and illiquid markets. Subsequently, increased allocation to bonds is associated with significantly lower investment fees, supporting the notion that investment fees are positively related to the riskiness of the assets held by the fund. Column 6 replaces hedge funds, private equity and infrastructure with the aggregate allocation to alternatives. Funds with higher allocation have higher investment fees, but the R-squared declines. The coefficients on the three individual alternative asset classes do differ in column 5 suggesting that not all alternative assets are the same. The positive asset risk-fee relationship observed from these results suggests that either the higher return potential of these assets justify the high associated fees, or simply riskier assets are more expensive to manage. Our results also confirm prior evidence in Australia that economies of scale exist, with larger funds having lower investment fees.

The most important conclusion from the analysis is that relating to fund type. The results show clearly that after one controls for the asset allocation decisions of the fund, that there are substantial differences in fees across fund types. In other words, if all funds were providing their members with the exact same asset allocation, there would be differences in fees. The results in column 5 of Table 5 show that corporate and public-sector funds provide investment fees that are 15 and 12 basis points lower than industry funds per annum, respectively. Retail funds, on the other hand, charge investment fees that are nine basis points higher than industry funds per annum. The evidence is clear that there are substantial differences in investment fees across fund types that are not driven by asset allocation or fund size. Presumably, part of the differences is driven by the profit margins that retail funds need to achieve. Superannuation fund members are paying more in investment fees for retail funds when ‘apples are compared with apples’. This suggests that the more expensive funds might have scope to lower their investment fees.

Table 6 contains the results when the administrative fee, converted to basis points for an account balance of $50,000, is used as the dependent variable. Before focusing our attention on column 5, column 1 shows that two-thirds of the variation in administration fee can be explained by fund type alone. Corporate and public-sector funds charge administrative fees on par with industry funds, yet retail funds charge an administration fee that is 60 basis points higher per annum, on average. This could reflect either the profit margin that retail funds need to earn or they could be overstated given the discounts on the ‘rack rate’ that retail funds provide to corporate-sponsored plans (Chant et al., 2014). The risk category variables do not help explain administration fees. We again observe economies of scale in administration fees, providing consistent evidence that larger funds do pass on lower fees to members. There is evidence suggesting that certain asset classes correlate with lower administrative fees, but this is only when we exclude the fund type variable. The independent variables in column 5 are able to explain over 71% of the variation in fund administration fees. The results show that retail funds charge significantly higher administration fees. Although we are unable to delve any deeper into the services provided to members in return for these higher fees, it is reasonable to expect the provision of substantial services. However, given that investors are primarily concerned with maximising the value of their retirement savings, subject to risk considerations, one can question whether there are any services that could be provided to justify the significantly higher costs.

5.3. Do high-fee funds earn higher returns?

Table 7 reports the average monthly performance of fee-sorted portfolios. We present raw returns, returns in excess of the risk-free rate and the abnormal returns relative to each fund’s AAB. Panel A contains the portfolios grouped by investment fee. In terms of raw returns, funds with low-investment fees generate an average raw return of 39 basis points per month and high-investment fee funds earn 42 basis points per month. This 3.6 basis points difference in monthly after-fee raw returns is about 43 basis points per annum. However, this difference is not statistically significant. Although the performance magnitudes change when we use excess returns and abnormal returns,¹⁹ the performance difference between the high and low-investment fee portfolios is similar. High-investment fee funds charge annual fees of 7.6 basis points per month, almost triple that of low-fee funds of 2.7 basis points per month. The difference in administration fee is small and economically insignificant. Panel B sorts the funds by administration fee. In terms of significance, the results are similar to those where portfolios are formed based on investment fee, as fees are not related to performance. Low administration fee funds generate slightly higher performance across the three metrics. Within the two panels, we also use after-fee return measures by subtracting fees from fund returns. For abnormal returns, we also need to consider subtracting fees from the benchmark returns that we use. To estimate fees for each benchmark, we use Mercer (2020) or Vanguard exchange-traded fund (ETF) fees that we show in the last column of Appendix 2. From Mercer (2020), where available, we select the median fee for each category for funds of $500m or larger. We find similar results to our gross fee measures. These results do not control for other determinants of performance, so that we now turn our attention to regression analysis to ascertain the drivers of fund performance.

Table 7.

Returns of fee-sorted portfolios.

Fee group	Before-fee performance			After-fee performance			Investment fee	Admin. fee
Fee group	Raw return	Excess return	Abnormal return	Raw return	Excess return	Abnormal return	Investment fee	Admin. fee
Panel A: investment-fee-sorted portfolios
Low	0.387*** (2.71)	0.057 (0.39)	0.272** (2.22)	0.325** (2.27)	–0.006 (–0.04)	0.226* (1.84)	0.027	0.036
2	0.392** (2.32)	0.062 (0.36)	0.281* (1.87)	0.316* (1.86)	–0.015 (–0.08)	0.220 (1.47)	0.041	0.038
3	0.418** (2.24)	0.088 (0.46)	0.293* (1.76)	0.328* (1.76)	–0.003 (–0.01)	0.222 (1.34)	0.051	0.039
4	0.412** (1.99)	0.082 (0.39)	0.284 (1.49)	0.313 (1.51)	–0.017 (–0.08)	0.203 (1.07)	0.059	0.042
High	0.423* (1.95)	0.093 (0.42)	0.310 (1.51)	0.311 (1.43)	–0.020 (–0.09)	0.221 (1.08)	0.076	0.035
High–low	0.036 (0.14)	0.036 (0.13)	0.039 (0.16)	–0.014 (–0.05)	–0.014 (–0.05)	–0.005 (–0.02)	0.049*** (65.43)	–0.001* (–1.85)
Panel B: administration-fee-sorted portfolios
Low	0.404** (2.33)	0.074 (0.41)	0.292* (1.84)	0.346** (2.00)	0.016 (0.09)	0.253 (1.60)	0.047	0.012
2	0.453*** (2.71)	0.123 (0.71)	0.316** (2.07)	0.384** (2.29)	0.053 (0.31)	0.267* (1.75)	0.050	0.019
3	0.405** (2.29)	0.074 (0.41)	0.312* (1.90)	0.323* (1.82)	–0.007 (–0.04)	0.244 (1.49)	0.057	0.027
4	0.394* (1.97)	0.063 (0.31)	0.259 (1.46)	0.301 (1.51)	–0.029 (–0.14)	0.185 (1.04)	0.047	0.046
High	0.366* (1.74)	0.036 (0.17)	0.257 (1.38)	0.235 (1.11)	–0.096 (–0.45)	0.146 (0.78)	0.053	0.079
High–low	–0.038 (–0.14)	–0.038 (–0.13)	–0.035 (–0.14)	–0.112 (–0.41)	–0.112 (–0.40)	–0.108 (–0.44)	0.007*** (8.63)	0.067*** (119.79)

Superannuation funds are sorted into quintiles according to their 1-month-lagged investment fee, over the period from June 2007 to March 2015, and are rebalanced monthly. Quintile one contains the lowest fee funds and quintile five are the highest fee funds. Performance is measured as the mean monthly raw fund returns, returns in excess of the risk-free rate (excess returns) and abnormal returns are measured as the fund-level monthly returns less the expected fund return based on the fund’s actual allocation benchmark (AAB). Average monthly investment and administration fees across the sample period are reported for each quintile in both panels. Portfolio returns and investment fees are equal-weighted across each quintile. The differences in performance and fees, between the high- and low-fee quintiles are also reported. t statistics are reported in parenthesis, and *, ** and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively.

5.4. What explains the performance of superannuation funds?

The results for the individual fund regressions are presented in Table 8. The dependent variables in these regressions are monthly fund returns in excess of the risk-free rate (columns 1–5) and the abnormal fund return relative to its AAB (columns 6–10). Columns 1, 2, 6 and 7 confirm the results from the portfolio sorts that monthly investment and monthly administrative fees are not related to either excess returns or abnormal returns of superannuation funds. Columns 3 and 8 include dummy variables for the different types of funds. Columns 4, 5, 9 and 10 add in the assets under management and a MySuper dummy variable as drivers of excess returns.

Table 8.

Fund performance and fee regressions.

	Excess returns					Abnormal returns
	1	2	3	4	5	6	7	8	9	10
Investment fee	1.168 (0.78)			1.276 (0.84)		1.201 (0.73)			1.087 (0.66)
Admin. fee		–0.435 (–0.69)		–0.225 (–0.36)			–0.714 (–1.25)		–0.752 (–1.24)
Retail			–0.044 (–0.84)		–0.038 (–0.75)			–0.071 (–1.45)		–0.073 (–1.51)
Corporate			0.026 (0.51)		0.022 (0.45)			0.053* (1.76)		0.048* (1.65)
Public sector			–0.009 (–0.33)		–0.009 (–0.35)			–0.013 (–0.43)		–0.011 (–0.37)
Size				0.018** (2.31)	0.012* (1.72)				–0.005 (–0.85)	–0.009 (–1.38)
MySuper				0.069* (1.94)	0.073* (1.92)				0.098* (1.92)	0.100* (1.88)
Adj. R²	0.819	0.819	0.819	0.819	0.819	0.778	0.778	0.778	0.778	0.778
Observations	10,910	10,910	10,910	10,910	10,910	10,910	10,910	10,910	10,910	10,910

Fund-level performance regressions are estimated for a sample of Australian superannuation funds over the period from June 2007 to March 2015 using monthly data. The dependent variable for columns 1–5 is the fund-level monthly returns in excess of the risk-free rate (excess returns). The dependent variable for columns 6–10 is the abnormal return, which is measured as fund-level monthly returns less the expected fund return based on the fund’s actual allocation benchmark (AAB). The independent variables are the monthly percentage investment fee, the monthly percentage administration fee and the log of assets under in management (Size). We include a series of dummy variables. Retail is equal to one if the fund is a retail fund and zero otherwise. Similar variables are included for corporate funds and public-sector funds. The industry fund group is omitted. We include a dummy that takes the value of one if the fund is a MySuper fund or zero otherwise (MySuper). Regressions are estimated with year-/month-fixed effects. All regressions are estimated with robust standard errors clustered by fund and year/month. t statistics are reported in parenthesis. *, ** and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively.

Our results reveal some particularly interesting differences in performance between fund types. The dummy variables for corporate, retail and public-sector measure fund performance relative to industry funds. When comparing the coefficients on these dummy variables in column 5, it is evident that there are no differences in performance when examining excess returns. However, when we take into account the asset allocation of each fund and adjust performance relative to this benchmark (AAB), we see performance differences arise. Column 10 shows that retail funds underperform industry funds, but this coefficient is not significantly different from zero. Corporate funds outperform industry funds and public-sector fund performance is similar to industry funds. However, the abnormal performance of retail funds is significantly worse than corporate funds and public-sector funds, being 12 and 6 basis points lower per month, respectively. Public-sector funds slightly underperform corporate funds.

MySuper is consistently significant at 10%, with these funds experiencing excess returns that are 7 basis points higher per month than non-MySuper funds, and abnormal returns that are 10 basis points higher. The MySuper funds only begin appearing in our sample in 2014. The finding of outperformance is particularly noteworthy given that MySuper funds were introduced as the default investment option to provide a low-cost product that would simplify choice for superannuation fund members. Our results show that this legislation has been a positive for superannuation fund members in these funds.

Another area of debate regarding Australian superannuation funds is whether small funds should merge, implying that small funds are not achieving economics of scale. Our results present mixed evidence on this front. In terms of excess returns, large fund options outperform. However, when we use our preferred measure of performance that adjusts for asset allocation, we do not observe any relationship with fund size and performance.

Overall, our results suggest that the drivers of outperformance are somewhat elusive. Although the explanatory power of the regressions is very high at around 80%, the number of significant factors impacting returns is low. This indicates that the time-fixed effects are explaining a lot of the variation in fund performance. It is clear that performance measures that have investment fees subtracted are not related to either measure of fees. So, this raises the question of what are investors receiving for their high fees? Clearly, it is not higher performance. But, on the other hand, it is not lower performance. In that respect, it seems that the market performance of these funds is relatively efficient. The only caveat to this statement is that fund type does help explain performance.

6. Conclusion

Using a comprehensive data set of for-profit, not-for-profit, default and non-default Australian superannuation fund options, this study shows that the investment fees charged by superannuation funds can be partially explained by the funds’ exposure to illiquid alternative asset classes, such as private equity and hedge funds that operate in complex markets. Given that the holding period of superannuation funds members is likely to be relatively long, it would appear reasonable for superannuation funds to attempt to benefit from diversification and any additional returns that may be available in these asset classes, for example, illiquidity premiums. After controlling for the asset allocations of superannuation funds, we find that the type of fund also has a significant influence on the investment fee charged. The conclusion we reach from the analysis is there seems to be an economic justification for the investment fees that are being charged and that calls to lower investment fees need to consider the risk tolerance of the investor and the asset allocation that they believe to be optimal. On the other hand, retail funds charge significantly higher fees than industry funds. The difference is in the order of 9–12 basis points per year, recalling that this estimate takes into account differences in asset allocation and fund size. These results suggest that certain superannuation funds would be able to lower investment fees to benefit their members. Our examination of administration fees reveals substantial differences across the type of fund.

We find that neither investment fees nor administration fees are related to fund performance net of investment fees. Adjustments for the risk-free rate and individual fund asset allocations do not impact the relationship between fees and performance. We do find that performance differences exist between certain fund types (e.g. retail and corporate funds, and retail and public-sector funds). MySuper funds outperform non-MySuper funds, supporting the legislative change that created these types of funds.

Overall, the results reveal two important findings regard fees and performance in superannuation funds. First, investment fees are driven by allocations to asset classes that trade in complex markets. However, these riskier allocations do not improve fund performance on a net-of-fee basis. Second, the type of fund impacts the fee and performance outcome of the superannuation fund member. Retail funds generally charge higher investment and administration fees than industry, corporate and public-sector funds. Retail funds also underperform compared to corporate and public-sector funds. At the other end of the spectrum, corporate funds charge the lowest fees and generate the highest performance. Any future regulatory changes need to consider these findings to improve the wealth accumulation outcome of superannuation fund members.

Footnotes

Appendix 1. Comparison of Chant West sample to APRA universe

Panel A. Comparison of Chant West sample to APRA sample
Year	Variable	For-profit	% of APRA	Not-for-profit	% of APRA	Total	% of APRA
2007	Net assets ($b)	140	32.95	145	57.25	285	42
2007	Number of entities	8	4.82	14	16.67	22	8.8
2008	Net assets ($b)	135	33.91	157	60.61	292	44.45
2008	Number of entities	8	5.59	17	21.52	25	11.26
2009	Net assets ($b)	149	41.89	159	64.4	308	51.11
2009	Number of entities	11	9.48	21	28	32	16.75
2010	Net assets ($b)	171	43.57	196	62.55	367	52.01
2010	Number of entities	11	10.68	24	32	35	19.66
2011	Net assets ($b)	185	43.89	271	74.42	456	58.04
2011	Number of entities	11	12.94	26	37.14	37	23.87
2012	Net assets ($b)	210	50.14	285	73.52	495	61.37
2012	Number of entities	12	15	26	39.39	38	26.03
2013	Net assets ($b)	238	49.75	316	68.01	554	58.74
2013	Number of entities	12	16.44	26	42.62	38	28.36
2014	Net assets ($b)	265	48.72	400	73.62	664	61.17
2014	Number of entities	12	17.91	26	45.61	38	30.65
2015	Net assets ($b)	288	49.23	454	73.44	741	61.67
2015	Number of entities	12	18.75	27	48.21	39	32.5
Panel B. Distribution of annual net assets ($b) of Chant West and APRA samples
	Mean	Median	Std dev.	10th percentile	25th percentile	75th percentile	90th percentile
Chant West sample
For-profit	18.35	13.18	17.49	2.24	6.15	19.27	48.95
Not-for-profit	11.51	6.00	13.93	1.12	2.92	14.25	29.17
APRA sample
For-profit	4.48	0.59	10.16	0.03	0.13	3.48	12.89
Not-for-profit	5.54	1.64	10.30	0.17	0.53	5.21	15.70

APRA: Australian Prudential Regulation Authority.

The tables report the nets assets and number of responsible entities represented in the Chant West sample to the APRA sample from ASIC’s annual fund-level superannuation statistics back series²⁰ by for-/not-for-profit status and by year to check whether the Chant West sample is representative of the superannuation fund universe. We find across years in our sample, coverage about half or more of net assets represented in the APRA sample by Chant West and at most a third of responsible entities by number. Coverage improves over time with later year having higher coverage both in net assets and number of entities. Across for-profit (corporate and retail funds) and not-for-profit (industry and public sector), the Chant West sample covers not for-profit the best at 76% of net assets and half of number of entities covered. For-profit fund coverage is at best 50% of net assets and 20% of number of entities. Part of the reason for the low representation in for-profit entities is the lack of coverage of corporate funds in Chant West. Overall, the Chant West has good representation across years and fund types by net assets and within the larger not-for-profit sector. Panel A shows our comparison of our Chant West sample to APRA’s. Panel B reports the distribution of net assets for the two samples.

Appendix 2. Benchmarks for each asset class

Asset class	Benchmark	Benchmark fee (% p.a.)/source
Australian shares	S&P/ASX 300 total return	0.53/Mercer (2020) survey
International shares	MSCI World Ex-Australia index hedged $A	0.06/Mercer (2020) survey
Australian property	S&P/ASX 200 A-REIT total return index	0.23/Mercer (2020) survey
International property	FTSE EPRA/NAREIT developed index hedged $A	0.40/International Property Securities Index Fund
Cash	RBA cash rate	0.06/Mercer (2020) survey
Australian bonds	Bloomberg AusBond composite bond index all maturities	0.26/Mercer (2020) survey
International bonds	Barclays global aggregate bond index hedged $A	0.07/Mercer (2020) survey
Private equity	AVCAL/Cambridge Associates private equity net return index	N/A
Infrastructure	Thomson Reuters global infrastructure total return index hedged $A	0.49/Vanguard Global Infrastructure Index Fund
Hedge funds	Eureka hedge fund index (equal-weighted net of fees)	N/A
Commodities	S&P GSCI total return index hedged $A	0.56/Mercer (2020) survey
Other alternative assets	Eureka hedge fund index (equal-weighted net of fees)	N/A

This table contains the benchmark indices used in the asset allocation benchmark (AAB) and factor model regressions, which capture the risks associated with the asset classes that are typically held by Australian superannuation funds. Furthermore, we report the benchmark fees that we use to calculate after-fee AAB. Hedge funds and other alternative assets are net fee, so that there is no benchmark fee.

Acknowledgements

The authors thank Chant West for the provision of data and Warren Chant, David Gallagher and Geoff Warren for helpful comments.

Final transcript accepted 23 October 2022 by Philip Gharghori (AE Finance).

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by Centre for International Finance and Regulation CIFR) grant T004.

ORCID iDs

Andrew Ainsworth

Adrian Lee

Terry Walter

Notes

References

Adams

Mansi

Nishikawa

(2012) Are mutual fund fees excessive? Journal of Banking & Finance 36: 2245–2259.

Alserda

Bikker

Van Der Lecq

(2018) X-efficiency and economies of scale in pension fund administration and investment. Applied Economics 50: 5164–5188.

Andonov

Bauer

Cremers

(2012) Can large pension funds beat the market? Asset allocation, market timing, security selection and the limits of liquidity. Network for Studies on Pensions, Aging and Retirement discussion paper 10/2012-062, October. Available at: https://www.netspar.nl//assets/uploads/062_Bauer.pdf

Australian Government (2020) Your Future, Your Super: Reforms to Make Your Super Work Harder for You. Canberra, ACT, Australia: Commonwealth of Australia.

Basu

Andrews

(2014) Asset allocation policy, returns and expenses of superannuation funds: Recent evidence based on default options. Australian Economic Review 47: 63–77.

Bateman

Mitchell

(2004) New evidence on pension plan design and administrative expenses: The Australian experience. Journal of Pension Economics & Finance 3: 63–76.

Bauer

RMMJ

Cremers

KJM

Frehen

RGP

(2010) Pension fund performance and costs: Small is beautiful. Working paper, Maastricht University, Maastricht, 30 April.

Bechmann

Rangvid

(2007) Rating mutual funds: Construction and information content of an investor-cost based rating of Danish mutual funds. Journal of Empirical Finance 14: 662–693.

Berkowitz

Kotowitz

(2002) Managerial quality and the structure of management expenses in the US mutual fund industry. International Review of Economics & Finance 11: 315–330.

10.

Bikker

(2017) Is there an optimal pension fund size? A scale-economy analysis of administrative and investment costs. In: Bikker

(ed.) Pension Fund Economics and Finance: Efficiency, Investments and Risk-Taking. London: Routledge, pp. 9–40.

11.

Bird

Liem

Thorp

(2014) Infrastructure: Real assets and real returns. European Financial Management 20: 802–824.

12.

Carhart

(1997) On persistence in mutual fund performance. The Journal of Finance 52: 57–82.

13.

Chant

Mohankumar

Warren

(2014) Mysuper: A new landscape for default superannuation funds. CIFR paper no. 020/2014, 1 April. Centre for International Finance and Regulation (CIFR), Sydney, Australia.

14.

Chant West (2014) A low fee diet is no guarantee of healthy returns. Available at: https://www.chantwest.com.au/resources/a-low-fee-diet-is-no-guarantee-of-healthy-returns/ (accessed 5 January 2022).

15.

Chen

Gallagher

Lee

(2017) Testing the effect of portfolio holdings disclosure in an environment absent of mandatory disclosure. Accounting and Finance 57: 101–116.

16.

Christoffersen

SEK

(2001) Why do money fund managers voluntarily waive their fees? The Journal of Finance 56: 1117–1140.

17.

Christoffersen

SEK

Musto

(2002) Demand curves and the pricing of money management. The Review of Financial Studies 15: 1499–1524.

18.

Coleman

Esho

Wong

(2006) The impact of agency costs on the investment performance of Australian pension funds. Journal of Pension Economics & Finance 5: 299–324.

19.

Commonwealth of Australia (2010a) Super System Review: Final Report, Part One – Overview and Recommendations. Canberra, ACT, Australia: Commonwealth of Australia.

20.

Commonwealth of Australia (2010b) Super System Review: Final Report, Part Two – Recommendation Packages. Canberra, ACT, Australia: Commonwealth of Australia.

21.

Commonwealth of Australia (2014) Financial System Inquiry: Final Report. Canberra, ACT, Australia: Commonwealth of Australia.

22.

Cooper

Halling

Lemmon

(2011) Fee dispersion and persistence in the mutual fund industry. Working paper, The University of Utah, Salt Lake City, UT.

23.

Cummings

(2016) Effect of fund size on the performance of Australian superannuation funds. Accounting and Finance 56: 695–725.

24.

Dellva

Olson

(1998) The relationship between mutual fund fees and expenses and their effects on performance. The Financial Review 33: 85–103.

25.

Drew

Stanford

(2003) Principal and agent problems in superannuation funds. Australian Economic Review 36: 98–107.

26.

Dyck

IJA

Pomorski

(2011) Is bigger better? Size and performance in pension plan management. Working paper 1690724, 1 June. Toronto, ON, Canada: Rotman School of Management.

27.

Ellis

Tobin

Tracey

(2008) Investment Performance, Asset Allocation, and Expenses of Large Superannuation Funds. Sydney, NSW, Australia: Australian Prudential Regulation Authority.

28.

Gallagher

(2001) Attribution of investment performance: An analysis of Australian pooled superannuation funds. Accounting and Finance 41: 41–62.

29.

Geranio

Zanotti

(2005) Can mutual funds characteristics explain fees? Journal of Multinational Financial Management 15: 354–376.

30.

Gil-Bazo

Ruiz-Verdú

(2009) The relation between price and performance in the mutual fund industry. The Journal of Finance 64: 2153–2183.

31.

Golec

(2003) Regulation and the rise in asset-based mutual fund management fees. Journal of Financial Research 26: 19–30.

32.

Gruber

(1996) Another puzzle: The growth in actively managed mutual funds. The Journal of Finance 51: 783–810.

33.

Guasoni

Obłój

(2016) The incentives of hedge fund fees and high-water marks. Mathematical Finance 26: 269–295.

34.

Haslem

Baker

Smith

(2008) Performance and characteristics of actively managed retail mutual funds with diverse expense ratios. Financial Services Review 17: 49–68.

35.

Higgs

Worthington

(2012) Economies of scale and scope in Australian superannuation (pension) funds. Pensions: An International Journal 17: 252–259.

36.

Iannotta

Navone

(2012) The cross-section of mutual fund fee dispersion. Journal of Banking & Finance 36: 846–856.

37.

Khorana

Servaes

Tufano

(2009) Mutual fund fees around the world. The Review of Financial Studies 22: 1279–1310.

38.

Korkeamaki

Smythe

(2004) Effects of market segmentation and bank concentration on mutual fund expenses and returns: Evidence from Finland. European Financial Management 10: 413–438.

39.

Kothari

Warner

(2001) Evaluating mutual fund performance. The Journal of Finance 56: 1985–2010.

40.

Lesseig

Long

Smythe

(2002) Gains to mutual fund sponsors offering multiple share class funds. The Journal of Financial Research 25: 81–98.

41.

Liu

Arnold

(2010) Australian Superannuation Outsourcing: Fees, Related Parties and Concentrated Markets. Sydney, NSW, Australia: Australian Prudential Regulation Authority.

42.

Luo

(2002) Mutual fund fee-setting, market structure and mark-ups. Economica 69: 245–271.

43.

Mahar

Mangnejo

Brohi

(2021) Determinants of mutual funds performance: A review article. Research Journal of Social Sciences & Economics Review 2: 61–67.

44.

Malhotra

McLeod

(1997) An empirical analysis of mutual fund expenses. The Journal of Financial Research 20: 175–190.

45.

Martínez Sedano

Gil-Bazo

(2004) The black box of mutual fund fees. Spanish Review of Financial Economics 4: 54–82.

46.

Mehri

Hassan

Safa

, et al. (2021) Do determinants of fees differ between Islamic and conventional funds? International Journal of Finance and Economics 26: 3599–3623.

47.

Mercer (2020) Global asset manager fee survey 2020. Available at: https://www.mercer.com/content/dam/mercer/attachments/global/gl-2020-global-asset-manager-fee-survey-2020.pdf (accessed 26 May 2022).

48.

Minifie

(2014a) Super savings. Grattan Institute report 2014-1. Available at: https://grattan.edu.au/wp-content/uploads/2015/04/821-super-savings.pdf

49.

Minifie

(2014b) Super sting: How to stop Australians paying too much for superannuation. Grattan Institute report 2014-6. Available at: https://grattan.edu.au/wp-content/uploads/2014/04/811-super-sting.pdf

50.

Nanda

Wang

Zheng

(2009) The ABCs of mutual funds: On the introduction of multiple share classes. Journal of Financial Intermediation 18: 329–361.

51.

Otten

Bams

(2002) European mutual fund performance. European Financial Management 8: 75–101.

52.

Phalippou

Rauch

Umber

(2018) Private equity portfolio company fees. Journal of Financial Economics 129: 559–585.

53.

Phillips

Pukthuanthong

Rau

(2018) Size does not matter: Diseconomies of scale in the mutual fund industry revisited. Journal of Banking & Finance 88: 357–365.

54.

Productivity Commission (2018) Superannuation: Assessing efficiency and competitiveness. Productivity Commission inquiry report overview, no. 91, 21 December. Melbourne, VIC, Australia: Productivity Commission.

55.

Rice Warner Actuaries (2014) Superannuation fees – Financial system inquiry. Rice Warner research report. Available at: https://www.ricewarner.com/superannuation-fees-financial-system-inquiry/

56.

Robinson

Sensoy

(2013) Do private equity fund managers earn their fees? Compensation, ownership, and cash flow performance. Review of Financial Studies 26: 2760–2797.

57.

Sialm

Starks

Zhang

(2015) Defined contribution pension plans: Sticky or discerning money? The Journal of Finance 70: 805–838.

58.

(2010) Cost, performance and portfolio composition of small pension funds in Australia. Journal of Pension Economics & Finance 9: 345–368.

59.

Tan

Cam

M-A

(2014) Does governance structure influence pension fund fees and costs? An examination of Australian not-for-profit superannuation funds. Australian Journal of Management 40: 114–134.

60.

Tufano

Sevick

(1997) Board structure and fee-setting in the US mutual fund industry. Journal of Financial Economics 46: 321–355.

61.

Wahal

Wang

(2011) Competition among mutual funds. Journal of Financial Economics 99: 40–59.