DB vs. DC investment performance and fee significance

The Center for Retirement Research at Boston College has released a Brief, Investment Returns: Defined Benefit vs. Defined Contribution Plans (the “BC Brief”), reviewing over 20 years of Form 5500 data to determine the rate of return in defined benefit and defined contribution plans. The report finds that, after accounting for size of plan and differences in asset allocation, DB plans outperform DC plans by around 0.7 percent, and that underperformance is “likely” due to the fees charged on DC investments.

In what follows we discuss and evaluate the BC Brief. Our conclusion: while BC’s comparison of DB and DC investment returns is provocative and raises some interesting issues, comparing DB plans and DC plans is like comparing apples and oranges. Unlike DB plans, DC plans are vastly more complicated and function based on fundamentally different objectives. Moreover, the comparison is, in one regard at least, fruitless: many sponsors have abandoned DB plans as a way of providing retirement benefits.

Understanding DC investment performance is, however, we believe, critical for plan sponsors committed to their participants’ retirement security. In a follow-uparticle we will consider the four factors that we believe are most significant: asset allocation; transaction costs; agency; and expertise.

The BC Brief

BC looked at 5500s from 1990 to 2012. It determined plans’ “rate of return” pursuant to a relatively straightforward formula. Rate of return equals:

Ending assets minus Beginning assets minus Cash flows

divided by

Beginning assets plus ½(Cash flows)

With this formula, BC could simply pull the numbers for the variables from 5500s and do the math.

Adjusting for asset allocation and fund size

BC then adjusted for asset allocation and fund size, factors other than fees that (in their view) might affect returns. We’re not going to go into detail on how BC did this adjustment. Bottom line: BC was working from 5500 data and a certain amount of educated-guessing was necessarily involved.

With respect to the period studied and the significance of asset allocation to returns, BC had an interesting observation:

Although the asset allocation of the two types of plans [DB and DC] differed significantly over the period 1990-2012, asset allocation would be expected to have only a modest effect on returns. The reason is that the long-run (1926-2014) pattern, where risky equities significantly out-performed less risky long-term corporate bonds, has not held over the past two decades ….

Our explanation for this would be that the pattern of declining interest rates from 1990 to the present has produced unusually favorable returns for bonds, compared with the entire period 1926-2014.

The significance of fees

Having accounted for the effects of plan size and asset allocation, BC concludes that the remaining DC underperformance is attributable to investment management fees:

[N]either size nor asset allocation is driving the differences in returns, which must be due to either differences in the performance of specific investments within the broader asset classes or, more likely, to investment fees.

Here’s BC’s table showing DB vs. DC performance after subtracting the effects of asset allocation:

Regression Results; Differential Between DB and DC Annual Returns; 1990-2012 and Sub-periods

Plan

1990-2012

1990-2002

2003-2012

All plans

Unweighted

0.7%

0.8%

0.3%

Weighted

1.2

0.9

1.4

Plans > $100 million

Unweighted

0.7

0.5

0.9

Weighted

1.1

0.7

1.5

Source: Center for Retirement Research at Boston College

“Weighted” returns are returns weighted for size of plan assets.

How big the difference revealed by BC’s analysis is depends on how you want to view the question – based on plans or based on assets, and whether you want to just look at larger plans.

Another interesting observation from BC: “Some researchers have suggested that the differential between defined benefit and defined contribution plan returns has declined over time, but the data show that the differential is generally larger after 2002 ….” So if there is a “problem,” that is, if DC plans are underperforming DB plans, it’s getting worse.

And finally, bigger size doesn’t always equal better performance: “for defined contribution plans … returns increase until plans reach $1 billion and then decline thereafter.”

Evaluating BC’s conclusion – why compare DB vs. DC returns?

In case it’s not clear: the whole point of doing this study is to quantify “how big a problem fees are in DC plans.” It’s fair to say that the BC Center for Retirement Research has a strong view that fees in DC plans are (1) too high and (2) reduce overall DC plan performance. This study and its conclusions clearly support that view.

Thus, this Brief must be viewed as a piece of advocacy. That doesn’t mean it’s wrong or flawed (for that reason). But it’s useful to understand that the primary point of BC’s comparison of DB and DC investment performance is to show that DC plans “underperform,” and DC fees are the major factor in that underperformance.

The comparison of DB and DC investment performance has a superficial appeal. As the story goes, most sponsors of DB plans have, over the last 20 or so years, frozen and/or terminated them and replaced them with DC plans. If it is the case – as the BC Brief purports to show – that DC plans do an inferior job of investing retirement assets, maybe that change was a mistake or, at least, maybe there is something “wrong” with DC plans that needs to be fixed.

But – are DB and DC plans actually “comparable?”

There are, we believe, three objections to the idea that comparing DC returns to DB returns is a good way of figuring out whether DC plans are doing a “good enough job.”

First, even though they are both “retirement plans,” DB plans aren’t really the same thing as DC plans. The conceptually difficult issue here is how to account for the employer guarantee in a DB plan – the employer’s obligation to make up for any performance shortfalls. Probably the best approach is to consider DB plan assets and liabilities and DB investment performance as “owned” by the employer. That is, indeed, what the accountants do: DB assets, liabilities, and investment performance are booked to the employer, not to the employee.

This approach views the true beneficiary of the DB trust as the employer, and the trust simply as a tool the employer uses to finance its liability to its employees – the pension promise it made. If we take this approach, however, the investment objectives that matter would be the employer’s, not the employees’. Thus, you might as well be comparing DC returns with returns on any other corporate investment, or the performance of endowments.

Second, DB plans – for most sponsors (and their employees) – aren’t really a viable alternative. As we just noted, in the private sector and with certain “special case” exceptions, employer commitment to DB plans is vanishing. And, as it happens, that’s entirely because of the sponsor guarantee. Thus, as a practical matter, there’s no point in comparing DB and DC returns, because, for many sponsors and their participants, you can’t get DB returns anymore.

Third, comparing (as BC does) returns at the plan level misses a fundamental difference between DB and DC plans. DB plans are, indeed, invested as one big account against one investment horizon with one overall risk/return objective. The overwhelming majority of DC plans are participant choice plans, where each individual participant invests against her individual investment horizon with her unique risk/return objective. It’s a methodological mistake to simply sum, on a plan by plan basis, all of those individual investments and call that the “rate of return” for the plan. We will discuss this issue further in our follow on article.

All of this is a very long-winded way of saying that DB plans are apples and DC plans are oranges.

Conclusion

We have not reviewed the raw data and methodology thoroughly enough to express an opinion as to whether BC’s regression analysis has adequately accounted for effects other than fees to explain the “underperformance” of DC plans. But, at a theoretical level, we believe it’s clear DC performance is a much more complicated issue than DB performance. There are many more things going on in DC plans: multiple investment horizons, more and different transaction costs, and different DC models – everything from brokerage windows to participants-choosing-mutual-funds to a TDF-based DC autopilot. In a DB plan you’re dealing with just one account.

Moreover, DC plans are invested with a view to the retirement savings objectives of (multiple and various) participants. DB plans are (in our view) invested based on the employer’s financial objectives.

Thus, we do not believe the BC Brief “proves” that DC fees are too high. That does not mean that there are not real issues with respect to DC fees. We do, indeed, believe that the issues raised by BC’s comparison of DB and DC investment performance are interesting and useful in trying to understand the weaknesses and strengths of the DC system. In our next article we will consider those weaknesses and strengths in four different dimensions: asset allocation; transaction costs; agency challenges; and expertise (or the lack thereof).