New Analysis of 401(k) plan performance and fees
Law professors Ian Ayres (Yale) and Quinn Curtis (University of Virginia) recently published a much-anticipated article – Beyond Diversification: The Pervasive Problem of Excessive Fees and ‘Dominated Funds’ in 401(k) Plans (John M. Olin Center for Studies in Law, Economics, and Public Policy Research Paper No. 493). The paper builds on work Ayres and Curtis have done on a database of more than 3,500 401(k) plans.
In the paper Professors Ayres and Curtis conclude that 401(k) plan participants suffer significant losses from (1) sponsor-fiduciary fund menu construction decisions, (2) participant asset allocation mistakes and (3) high fees on plan investment options. To remedy this problem they make several innovative proposals for changes in the rules for 401(k) plan fund menu construction.
In this article we review their paper in detail.
In this article we are going to treat Ayres and Curtis’s methodology as both valid and robust and then consider their findings and the policy proposals that flow from them. In our next article we will consider the limitations of and problems with this methodology.
Ayres and Curtis begin with a proprietary database of 401(k) plans constructed by Brightscope, Inc. that includes information about plan administration and expenses, the menu of investment options offered by the plan, and the balance of plan funds invested in each option.
From the Brightscope database, Ayres and Curtis use only “plans that exclusively utilize publicly listed mutual funds for their risky investment options.” This allows them to identify – from publicly available mutual fund prospectuses – the expense ratio for these funds in each of the plans they are observing. As they note, this limitation excludes a significant number (71% of the Brightscope database) of larger plans that use collective trusts or separate accounts. The result is a database that is (somewhat) biased away from larger plans.
Ayres and Curtis’s data is aggregate plan data. That is, for instance, with respect to Plan X they know that 23% of plan assets are invested in Fund Y; they do not know individual participant allocations to Fund Y.
To measure the performance of the plans they are considering, Ayres and Curtis compare:
1. The performance of an ideal, low-cost portfolio not restricted by the choices in any particular plan menu (the “after-fee return that was achievable in the market by investing in a low-cost well-diversified plans (sic)”). This ideal portfolio is based on an allocation between (1) the Russell 3000, (2) the Barclay’s US Aggregate Bond index, and (3) the MSCI EAFE international equity index.
2. The performance of an optimal portfolio constructed from the choices actually available in the plan menu.
3. The performance of the actual portfolio the plan held (again – this is an aggregate number, not the portfolio of any given participant).
Obviously, plan portfolios have more complicated asset allocations than the 3-asset (Russell 3000, Barclay’s Aggregate and MSCI EAFE) portfolio in the ideal portfolio. To deal with this – to get everything on an apples-to-apples basis – Ayres and Curtis “adjust the portfolios for each plan so that they all produce the same level of expected financial risk.”
Performance in each case is expected performance relative to risk – that is, expected risk-adjusted return. Expected returns are estimated based on stocks and bonds performance (Russell 3000, Barclay’s Aggregate and MSCI EAFE) over the period 1980-2000 (performance for the period 2001-2009 was considered too idiosyncratic to use as a baseline).
With this approach, ‘losses’ can then be defined as risk-adjusted returns that are lower than the risk-adjusted return of the ideal portfolio. Risk-adjusted returns are determined using Sharpe Ratios, which we understand to be (1) expected returns in excess of a risk-free return, divided by (2) the standard deviation of the return (representing volatility).
Generally, Ayres and Curtis exclude from their analysis investments in brokerage windows and guaranteed investment contracts. They do, generally, include company stock.
With this methodology, Ayres and Curtis can measure the following:
Menu Diversification Losses. These are the losses that arise because 401(k) investors are restricted to the menu of funds offered by their plan.
Menu Excess Expense Losses. This measures losses due to excess plan administrative fees and excess investment management fees charged by the mutual funds. [To determine “excess plan administrative fees and excess investment management fees” Ayres and Curtis compare actual plan costs to the cost of retail index fund fees (25 basis points) and low-cost plan administration fees (8 basis points).]
Investor Diversification Losses. This reflects the differences in returns that investors could receive on the optimal portfolio for their plan, and the returns investors are actually expected to receive based on the actual portfolio held by plan investors. This measure captures, with important caveats, mistakes such as under-diversification that are unrelated to fees. [The caveat – that the data reviewed is only aggregate data and does not reflect actual participant allocations/allocation mistakes – significantly compromises the learning here.]
Investor Excess Expense Losses. This measures difference between fee losses on the optimal post-fee portfolio and fee losses on the actual portfolio investors hold. It reflects fees that are incurred from investors deviating from the optimal menu portfolio.
Ayres and Curtis’s (fairly ambitious) goal is to identify where 401(k) plan participants are losing money. The following are their key findings:
The menus offered in 401(k) plans generally allow adequate diversification. The average loss participants sustain because they are unable (given plan choices) to adequately diversify is 6 basis points.
Menu excess expense losses are significant. In order to invest in an optimal portfolio, participants on average have to incur 43 basis points in excess fees (remember, excess fees are defined as fees above what would be paid in a retail index fund [generally 25 basis points]).
Participant asset allocation mistakes also trigger significant losses – averaging 65 basis points.
Finally, participant driven excess fee losses – fees incurred that would not have been incurred if the participant had selected the optimal portfolio in the plan – are also significant – averaging 49 basis points.
One of the important issues that the Ayres and Curtis paper identifies is the additional cost – in higher fees/lost returns – resulting from the inclusion of a ‘bad fund’ in a menu. This issue is important because, as Ayres and Curtis characterize it, some courts are adopting a rule that “[w]hile a menu that offers only poor options may be legally deficient, a menu that offers at least some good options will benefit from the protection of the [ERISA section 404(c)] safe harbor.”
Ayres and Curtis challenge the rationale behind this rule that, so long as there are ‘good’ choices available to participants, it is permissible to include some ‘bad’ choices in a fund menu. They believe that the inclusion of a bad fund option will generally result in participant losses. In support of this claim they cite:
[T]tendencies identified by behavioral economists Benartzi and Thaler [Shlomo Benartzi & Richard H. Thaler, Naive Diversification Strategies in Defined Contribution Saving Plans, 91 AMER. ECON. REV. 79â€“98 (2001)], for example, [finding] that investors tend to follow a naïve diversification strategy of allocating their funds equally across options on the menu, even when such allocation is not consistent with their investing goals.
To address this problem they propose singling out for special treatment what they call ‘Dominated Funds.’ They identify ‘Dominated Funds’ as “fund options where the costs of fees in holding the fund so outweigh the benefits of additional diversification that rational investors would not invest in these assets.” More technically, they would treat a fund as ‘dominated’ if either:
1. (a) There is another fund offered in the same plan menu of the same investing style [using 82 Morningstar style categories and whether the fund is an index fund] as the candidate fund with fees at least 50 basis points lower, and (b) the candidate fund has fees 25 basis points higher than the mean fees of funds with the same investing style in [their 3,500 plan sample], and (c) the candidate fund receives less than 1% weight in our computation of the optimal portfolio for the plan.
2. (a) There is no other fund in plan menu with the same style, and (b) the candidate fund has fees 50 basis points higher than the mean fees of funds with the same investing style in [their 3,500 plan sample], and (c) the fund receives less than 1% weight in our computation of the optimal portfolio for the plan.
The details of this technical definition are not as important as the basic idea: it is possible (Ayres and Curtis believe) to objectively identify funds in a plan’s fund menu the costs of which (in fees) so outweigh the benefits (in additional diversification) that they deserve special treatment. A key feature of their paper is the special treatment they propose for Dominated Funds.
Problems with the fiduciary standard for fund menu construction
Ayres and Curtis identify two problems with the current ERISA fiduciary standard for fund menu construction:
1. The Hecker-style cases allow the inclusion in plan fund menus of Dominated Funds, and the naïve participant diversification bias results in some (meaningful number of) participants selecting those ‘bad’ funds.
This problem is exacerbated by a ‘double-agency’ problem: the employer-fiduciary (rather than the participant himself) is selecting the fund menu that the participant must choose from. And this fund menu design decision is often made based on advice from a consultant who may have conflicts (e.g., may be in part compensated by the funds included in the menu). In this situation, the advisor (the consultant) is biased and the decider (the sponsor-fiduciary) has no stake in the decision, making the inclusion of a Dominated Fund easier.
2. Courts generally focus on prudent process rather than prudent results. This focus may be appropriate with respect to fund performance, as performance can only be evaluated retrospectively. But Ayres and Curtis believe this approach is unjustifiable with respect to fees, which are known in advance.
Policy proposal – a new ERISA fiduciary standard for fund menu construction
To address these problems, Ayres and Curtis propose a new fiduciary standard for fund menu construction:
In evaluating plan menus when excessive fees are alleged, courts ought to ask, in light of other options in the plan menu, could a prudent person reasonably believe that the fund in question ought to be held by investors?
At the risk of belaboring the obvious: this is a substantive standard, not a process standard. Under it the inclusion of a Dominated Fund in a fund menu would not be a per se breach of fiduciary duty, but the fiduciary would have to provide an adequate defense of the adoption of the high-fee fund – a reason why a prudent person would have bought the fund.
Effect on revenue sharing
Ayres and Curtis believe that the adoption of this new standard will make revenue sharing arrangements more difficult to implement. They believe that is a good thing:
[T]hese reforms may … result in costs that were previously covered out of revenue sharing [paid for, e.g., out of fees on a Dominated Fund] being billed directly to the plan. We view this [as] a positive change for two reasons. First, revenue sharing creates perverse incentives for plan advisors to suggest dominated funds. Second, revenue sharing promotes a pernicious cross-subsidization of plan-level expenses from less sophisticated to more sophisticated investor/employees.
Policy proposal – biasing participant choices towards low-cost investments
Ayres and Curtis would, however, go further than simply tweaking the fiduciary standard for fund menu construction. They advocate the adoption of rules that would ‘steer’ participants to invest in low-cost funds.
Mandate that the default fund be an EQDIA
Ayres and Curtis would establish a new fund category, the Enhanced Qualified Default Investment Alternative. An EQDIA would have to meet the requirements of a QDIA (e.g., a target date fund, a balanced fund, or a managed portfolio) and have expenses of less than 50 basis points. In this regard, they observe: “[o]ur proposal would not prohibit actively managed funds, and actively managed funds are available to institutional investors at this price point, but as a practical matter, most EQDIAs will be passively managed.”
Participants would have to pass a test to opt out of the default investment
Ayres and Curtis propose the adoption of ‘altering rules’ (‘sticky defaults’) that would limit the ability of participants to opt out of the EQDIA default. Generally, participants who want to opt out would have to pass a Department of Labor-developed ‘401(k) Investment Sophistication’ test. Alternatively, they would “permit employees acting pursuant to professional financial advice from, for example, a certified financial planner, to bypass the testing requirement.”
Policy proposal – require ‘high cost’ plans to allow in-service IRA rollovers
Finally, Ayres and Curtis propose:
If the average total percentage costs paid by all investors in a plan exceed a regulatory threshold, then investors in the plan should be permitted to roll over on a continuing basis their investments in the plan, without penalty, into an individual retirement account offering qualified, low-cost investments.
They would identify as ‘high cost’ any plan “with average plan and fund level costs that exceed the average expense ratios of a mixed portfolio of index funds by 125 basis points.”
‘Soft’ or ‘libertarian’ paternalism
These proposals are, obviously, quite radical and at odds with the spirit that at one time informed 401(k) saving and investment – freedom of participant choice. That spirit was, of course, never a reality – the sponsor-fiduciary chose the fund menu and providers. And it has been steadily eroding under criticism by, e.g., behavioral economists, who advocate ‘framing’ and ‘steering’ choices and by participant advocate groups and some policymakers who simply want better results.
Ayres and Curtis characterize their approach as a ‘soft’ or ‘libertarian’ paternalism – they cite Thaler and Sunstein’s Nudge four times. They argue this approach is justified by their own data – average losses in 401(k) plans of 156 basis points because of bad sponsor-fiduciary and participant choices – and by the now-familiar literature on the financial ignorance of participants.
* * *
Few would deny that there are inefficiencies in the current 401(k) system. Returns for at least some participants are lower for all the reasons Ayres and Curtis identify: sponsor-fiduciary fund menu construction decisions; participant asset allocation mistakes; and high fees on plan investment options.
The Ayres and Curtis paper is an important effort to try to put some numbers on the resulting losses and to identify which are the big problems. Their broad conclusion is that diversification at the plan level is less of a problem than (1) fees and (2) individual diversification decisions made by participants. To address those issues they propose: significant changes in the general fiduciary standard for fund menu construction; a set of ‘sticky defaults’ that would aggressively steer participants towards a low-cost QDIA; and mandating an option for participants in high-cost plans to rollover in-service.
There are some obvious issues with Ayres and Curtis’s methodology, and their proposals would disrupt the current 401(k) plan system. In our next article we evaluate both their methodology and their proposals.