Factors affecting DC returns
We recently discussed the Boston College Center for Retirement Research Brief, Investment Returns: Defined Benefit vs. Defined Contribution Plans (the “BC Brief”). The Brief reviewed over 20 years of Form 5500 data and determined that (after accounting for asset allocation and plan size) DB plans outperform DC plans by 0.7 percent. As we said in our article, while BC’s comparison of DB and DC investment returns is provocative and raises some interesting issues, DB plans and DC plans are apples and oranges.
Understanding DC investment performance is, however, we believe, critical for plan sponsors committed to their participants’ retirement security. In this article we consider the four factors that we believe are most significant: asset allocation; transaction costs; agency challenges; and expertise (or the lack thereof).
Asset allocation
Let’s compare two types of DC plans:
Plan A has a balance fund invested against an investment horizon that is the sum of the horizons of its participants. Thus it looks a lot like a DB plan minus the employer guarantee. There are very few DC plans that operate like this.
In Plan B, each participant’s account is invested in a way suited to that participant’s investment objectives. This could be done by letting the participant choose from an investment fund menu or simply by putting each participant in an appropriate target date fund (TDF). The overwhelming majority of DC plans allow, in one way or another, this sort individualization of investments.
We believe that, all other things being equal, it is axiomatic that Plan B does a better job of asset allocation, because allocations under it are tailored to individual investment horizons. And – although not particularly intuitive – that axiom holds even where Plan A’s investment return (at the plan level, as reported on its 5500) is higher than Plan B’s.
Transaction costs
We also think that it is axiomatic that Plan B will have higher transaction costs. In Plan A you have only one account, one pile of assets to manage. In Plan B you have an account for every participant.
The (axiomatic) gains from individualized asset allocation against individual investment horizons offset the (axiomatic) additional transaction costs. At the margin, if the costs are less than the benefits, there will be more individualization; if the costs are greater than the benefits, there will be less. We would claim that the information revolution has reduced the cost of individualization across a wide variety of domains. At this point it would be shocking if employers of any size could not efficiently individualize investment.
The ongoing challenge to DC sponsors is how to maximize the gains from individualization and minimize the losses to transaction costs.
Agency
DC plans suffer from an agency problem. Primary investment decisions – e.g., about what funds to put in a fund menu or which TDFs to use – are made by the plan sponsor. But the person who ultimately is affected by those choices – who experiences the gains or losses on plan investments – is the DC plan participant. Some sponsors (at least), in making those primary investment decisions, may be motivated by concerns other than optimization. For instance, by concerns about “getting sued” or about their relationship with their broker. Or the sponsor may simply be under-motivated.
DB plans do not suffer from this problem, because of the sponsor guarantee. That guarantee lines up sponsor incentives with investment performance. As we just said, no such alignment exists in DC plans.
This non-alignment of incentives has been a DC plan issue since their inception. Our tools to deal with it are crude: legal “rules” (most obviously ERISA’s fiduciary rules); and exhortation. The only “cure” is to put a brokerage window in every plan, because in a brokerage window the participant can make her own primary investment decisions. That cure, however, may be worse than the disease. Because, while a brokerage window may solve the agency problem, it may also exacerbate losses to transaction costs and (lack of) expertise.
Expertise
If we are reading the BC Brief correctly, an even greater problem for retirement investing generally is the ignorance and gullibility of participants (and in some cases plan sponsors). BC found that, over the period 2000-2012, “IRAs produced substantially lower returns than defined contribution or defined benefit plans.” In their view, that result was due to owners of IRAs being sold “high-fee products.”
Expertise/lack of expertise affects returns in DC plans in at least a couple of ways. First, to the extent that the sponsor is making the “primary” investment decision (constructing the fund menu or selecting TDFs), sponsor lack of expertise may result in “bad choices for the participant to choose from.” And, second, the participant may himself make bad choices – an inappropriate asset allocation or an unreasonably high-fee fund.
Expertise is certainly something that sponsors can do something about, by educating themselves and their participants, and by structuring participant choices (e.g., through defaults to TDFs) to produce better outcomes.
What has any of this got to do with fees?
As we said in our prior article, one of the main points (if not the primary point) of the BC Brief is that DC plans under-perform DB plans because of higher fees paid for DC investment management. All of the factors we’ve been discussing can affect fees.
A more complicated plan design or fund menu, more robust participant services, use of mutual funds (e.g., “because participants want to read about them in the paper”) all create transaction costs that ultimately translate into higher fees. As we said above, one sponsor objective should be to reduce transaction costs. There is, however, always a tradeoff between, e.g., “quality of service” and fees. A good question is, however, to what extent will more expensive, “higher quality” service translate into higher returns? If it does not, what is the justification for it?
The connection between agency issues (the non-alignment of interests) and lack of expertise with high fees is both obvious and intuitive.
And how – ultimately – do these factors affect returns?
Let’s go back to that “non-intuitive” point we made above. The individualization of investments in a DC plan does a better job of asset allocation, and that is true even where returns at the plan level for the DC plan (in our example, Plan B) are lower than returns on the single fund.
Let’s illustrate this. Assume our plan has four participants, one aged 35 and three aged 65, each with a $10,000 account. And assume that the optimum asset allocation for the 35 year old is 100% equities and for the 65 year olds 100% fixed income. In Plan A (with the balance fund) all accounts are invested 3/4 equity 1/4 fixed income. In Plan B each account is invested at the optimum asset allocation. In Year 1, equities lose 10%; fixed income earns 2%. In Plan A, the return is -1%. In Plan B, the plan level return (summing the results for all four participants) is -1.33%, which is 33% worse than Plan A. But at the participant level, in Plan A the 65 year old is clearly taking too much risk and the 35 year olds are not taking enough risk. In other words, the balance fund (or, dare we say, a DB plan) “performs” better at the plan level, but the individualized fund (DC plan) does a better job of asset allocation at the participant level.
All of this is to say that studying DC returns based on plan level Form 5500 data begs some pretty important questions.
What sort of DC performance metrics might make sense?
We are left with two pretty big issues. First – how do we evaluate whether DC assets are underperforming? What exactly should we be comparing DC returns to?
We’ve described our objections to comparing them to DB returns. Comparing them to IRAs makes, from one viewpoint, more sense; IRAs are, at this point, the only realistic retirement savings alternative available, in the US at least. But IRAs are kind of a low bar – the challenges with respect to transactions costs, agency and expertise are all greater for IRAs than for the typical DC plan.
Alternatively, should we simply compare them to the returns on an all-passive fund? That view is implicit in much of the “401(k) fees are too high” literature (consider, e.g., our article 401(k) plans and high fee investment funds). Perhaps – but in our view those analyses – which typically focus on S&P 500 returns – don’t adequately account for the benefits of diversification into investments that are more difficult to “passively” invest in.
So one critical theoretical task should be to develop adequate benchmarks that account for more than just the performance of US large cap funds.
Second, how do we methodologically account for the individualization of DC plan investments? It’s conceivable that a given plan does better for equity investments or younger participants or participants with outside assets or passive-only investors, and not so well for other participants.
It would be incredibly useful to get better participant level data. That data could then be used to grade plans at the participant level and then sum those grades for an overall plan level grade.
* * *
In our view, the purpose of the project of evaluating retirement investment returns should not be to identify which retirement savings vehicle is “worse” but to identify where there are problems that can be solved. Hopefully we will be able to develop methods and data that sponsors can use to do that.