Reconciling the 401k Fee Estimates of the ICI and its Critics

fee-study

Click Image to View the Study.

In April, the Investment Company Institute (ICI) released a 401(k) fee survey entitled “Inside the Structure of Defined Contribution/401(k) Plan Fees: A Study Assessing the Mechanics of What Drives the ‘All-In’ Fee.”  ICI chairman John Murphy* has begun using the median fee from this survey  – 0.72% – as evidence that fees in the entire 401k marketplace are “quite a bargain compared to the 3% some critics cite.”  To view how Mr. Murphy is using the data from this survey please see “401k Fee Critics Overstate Problem” published last week by Sara Hansard of Investment News. Our question is simple: how can the “All-In Fees” calculated by ICI (0.72%) and those calculated by its critics** (3.00%) be 2.28% apart? We propose that this discrepancy exists because of poor sampling methods in the ICI study as well as major differences in how fees are calculated by ICI and its critics.

Sampling Problems in the ICI/Deloitte Survey

When conducting any survey it is of utmost importance to make sure the survey sample is representative of the entire population. Otherwise the survey’s results are useless (or at best only representative of the sample and not the population as a whole).  ICI/Deloitte argues that they believe the demographics of the sample “appear to be similar” to the DC plan population and therefore the results are “representative.” We intend to use ICI/Deloitte’s own logic on what drives fees to show that the sample misrepresents the DC plan population on the two most important metrics: asset size and participant size.  We compared the ICI/Deloitte sample to the entire defined contribution marketplace based on data pulled from Form 5500 filings for the 2006 plan year:

1. Asset Size: The ICI/Deloitte sample contains a larger percentage of large asset plans than the DC plan population as a whole:

plans-by-asset-size-chart1

plans-by-asset-size-data

Although roughly 96% of plans in the entire DC market are micro/small plans (as defined by ICI/Deloitte), plans this size only make up 20% of the  ICI/Deloitte sample. Because plan asset size is identified by ICI/Deloitte as the “primary driver of fees” having such a grossly misrepresentative sample would lead the ICI/Deloitte study to systematically under-estimate fees for the median plan in the DC marketplace.

2. Participant Size: The ICI/Deloitte sample contains a larger percentage of plans with 1,000 or more participants than exists in the DC marketplace:

plans-by-participant-size-chart1

plans-by-participant-size-data

Although less than 1% of plans in the DC marketplace have 1,000+ participants, these plans make up a full two thirds of the ICI/Deloitte sample.  As described in the study, as the “number of participants rise, fees as a percentage of assets tend to fall.”  If this is indeed the case, we can expect that the ICI/Deloitte sample systematically underestimates fees for the median 401k plan.

Due to the inadequacy of the sample in the two areas deemed most critical in driving fees by ICI/Deloitte – asset size and participant size – it is safe to say that the survey systematically underestimates median fees for the DC marketplace. However, because ICI/Deloitte does not disclose its sampling methods it is impossible to know what caused such a dramatically misrepresentative sample. We hope that the poor sample is the result of self-selection bias rather than something more sinister. After all, it is entirely possible that such a sample could have resulted from a voluntary simple random sample. With over 1,000 data elements gathered, the survey was likely difficult and time consuming to complete. This complexity might have discouraged small plan sponsors from completing the survey. As a result the survey design may have involuntarily led to steep self-selection bias and an undercoverage of the small plan sponsors which are most likely to have plans with higher fees. While we believe that the situation we just described – that of self-selection bias and a clear lack of sampling rigor – is the most likely cause of the sampling problem, several facts point towards deeper problems.  We are troubled that both sources of bias (asset size and participants) cause underestimation of fees, that in spite of obvious evidence to the contrary ICI/Deloitte describes their results as “representative,” and, perhaps worst of all, the survey’s sampling methods remain completely undisclosed. Faced with these three facts, inquiring minds wonder if the survey’s designers were either mildly disingenuous or worse yet, lacking in intellectual integrity.

Reconciliation #1:  Fortunately we can use some of ICI/Deloitte’s own data to help them overcome their poor sampling.  In the study the authors include a chart of the median fee broken out by asset size.  The median fee for micro plans is listed as 1.89%.  Since micro plans make up two thirds of the entire DC marketplace it is safe to assume that the median all-in fee for all DC plans is very close to this 1.89%.  If the ICI intends to share a single fee number that is representative of the 401k marketplace according to its own calculations we recommend it use 1.89% .  However, this 1.89% still leaves another 1.11% (3.00% – 1.89%) gap between ICI and its critics.

What is included in the “All-In” Fee?

The rest of the gap between ICI and its critics can be closed by examining what is included in the “All-In Fee.”  It turns out that many critics – independent fiduciaries, consultants and RIAs – include transaction costs in their calculations while the ICI does not. The critics argue that the Department of Labor makes it very clear that fiduciaries have a duty to calculate and understand all fees of an investment alternative that lower returns, including the option’s transaction*** costs – brokerage commissions, spread costs, market impact costs and opportunity costs. Recent academic research and data from the SEC estimates that transaction costs are in many cases as big or bigger than a fund’s expense ratio.****

These transaction costs are real costs acknowledged by the ICI.  The ICI does not argue that transaction costs are not real nor does it argue that they are small. In fact, the ICI agrees that these fees are real and suggests that the turnover rate is a good proxy for estimating these expenses.  As evidence of this, the ICI recently released a list of “Myths about 401k Plans.”  Myth #7 addressed transaction costs:

ICI Myth #7: The cost of 401(k)s invested in mutual funds is substantially understated because funds don’t disclose trading costs – a hidden and excessive fee.

ICI Fact #7: Funds follow SEC rules on disclosing trading costs – and fund managers have strong legal and market incentives to minimize those costs.

The most interesting part of the ICI’s response to myth #7 is that they don’t actually refute the two most prominent pieces of the “myth”: that transactions costs are hidden and excessive. Their response – that funds follow SEC rules, and that fund managers have incentives to keep these costs low – does not refute the existence of these costs or the academic research that suggests that these costs can be as big (or bigger) than a fund’s expense ratio. In fact, the ICI’s only complaint is that these costs are hard to measure. Unfortunately for an ERISA fiduciary, difficulty in measurement does not relieve them of their fiduciary obligations to ensure reasonableness of fees.*****

Reconciliation #2: If a fiduciary follows the ICI’s advice and uses the turnover ratio to estimate these transaction costs, and then adds this fee in to a plan’s “All-In” fee the result is a number very similar to what the ICI’s critics allege.  ICI’s survey says that roughly 74% of plan fees in its survey are investment management fees.  We can use this math to determine that 1.4% (0.74 * 1.89%) of the 1.89% median all-in cost for micro plans in ICI’s survey are investment costs. Assuming transaction costs are roughly equal to investment costs, we can add another 1.4% for transaction costs and the result is 3.29% (1.89% + 1.4%).  This new number including transaction costs is actually higher than the 3% number ICI ascribes to its critics.  While this is a very crude calculation, it shows that the ICI and its critics are not in fact all that far apart if everyone can agree on what to include.

Adding Nuance to the Fee Debate

The ICI would like to claim that its critics are out in left field on fees as I’m sure the ICI’s critics would like everyone to believe that the ICI is out in right field.  The reality is that the two aren’t that far apart.  All that is needed to close the gap is a dose of reality about what the real DC plan population looks like and a little bit of nuance in understanding differences in how 401k “All-In Fees” are calculated.  From our perspective, there are three major take-aways from this analysis:

  1. Transaction Costs Are Real Fees: A strict reading of ERISA law and the DOL’s interpretive guidance makes it very clear that transaction costs must be measured and included in a fiduciaries analysis of 401k plan fees.  Until the DOL says anything to the contrary we believe prudent fiduciaries should include these fees in their calculations.
  2. Always Disclose your Sampling Methods: The ICI should do a better job of disclosing its survey and sampling methods.  After all, it is up to the readers to determine if the results of a survey are representative not the sponsors of the survey (especially when the sponsors are so obviously conflicted).  In this case we believe the current median number used by ICI in the press – 0.72%- is misleading and detrimental to their own credibility, especially since they claim directly in the survey report that the results “cannot be projected to the entire population of U.S. 401k plans.”
  3. Fee Debate: Both the ICI and their critics should be more nuanced in their debate on 401k fees. We would like to see more discussion of transaction costs and more rigorous sampling techniques for fee studies. For our part we believe we are assembling an amazing database of real plan fee data (not survey data) that will help everyone in the industry better benchmark 401k plans.

Call to Action

We call on the ICI to immediately release more information about its survey methods and explain why they think this current survey is representative of the entire 401k marketplace.

Disclosures:

* I believe it is important to point out that Mr. Murphy is also the chairman of Oppenheimer Funds, one of the nation’s largest mutual fund families.

** While Mr. Murphy didn’t mention anyone by name, independent fiduciary (and BrightScope advisory board member) Matthew Hutcheson did mention 3% in his article published in the Elder Law Journal entitled “Uncovering and Understanding Hidden Fees in Qualified Retirement Plans” published in the fall of 2007: “I have consistently found that low-cost plans cost 3% of plan asset annually.”

*** Transaction costs are also called trading costs or implicit costs.

****For a full analysis of this topic and links to academic research please see the SEC’s concept release on transaction costs.

***** Kasten, Gregory W., “High Transaction Costs from Portfolio Turnover Negatively Affect 401(k) Participants and Increase Plan Sponsor Fiduciary Liability”: “The
fact that the turnover costs are “hard to find” does not give the plan fiduciary the leisure of deciding not to monitor them.”

Leave a Reply