Towers Watson has been comparing investment rates of return in defined benefit (DB) and defined contribution (DC) plans for more than 10 years, 1 and DB plans have been the long-term victor. This analysis updates our prior studies with investment returns for 2006 2 and 2007 for a large set of plans, as well as a snapshot of year-end returns for 2008 based on a small set of plan sponsors.
In our last analysis, we found that between 1995 and 2006, DB plans outperformed DC plans by an average of 1 percentage point per year. Earlier studies also found that, over time, DB plans attained higher returns than 401(k) plans. In this year's analysis, the results remain in line with past analyses; DB plans outperform DC plans by roughly an average of 1 percentage point a year.
Our current analysis of year-end 2008 also shows — albeit drawn from a reduced sample — that despite generally poor returns for both plan types during the financial crisis, median returns for DB plans remained around 1 percentage point higher than those for DC plans — and some DB plans even reported positive returns.
Like our earlier studies, this analysis is based on Form 5500 financial and pension disclosure data through 2007 released by the Department of Labor (DOL). Results for 2008 are based on a Towers Watson survey of plan sponsors about their Form 5500 information. We also use the same formula as earlier studies to calculate the average rate of return:
Comparison for sponsors of both plan types, 1995-2007
We initially include only companies that sponsored one DB plan and one 401(k) plan, each with at least 100 participants. We limit our analysis to these companies to minimize the effects on the results of specific company or workforce characteristics uniquely associated with the sponsorship of only one plan type. This approach enables us to concentrate on differences in rates of return more directly associated with retirement plan type. 3
Plan size still makes a difference
Last year, we began measuring median rates of return weighted by plan asset size. The weighted basis is more relevant to understanding the experience of plan participants because, after all, plans with more assets generally have more participants. Figure 1 compares rates of return in DB and 401(k) plans from 1995 to 2007, weighted by plan asset size.
Figure 1. Asset-weighted median rates of return for DB and 401(k) plans, 1995 to 2007*
* Data for years before 2006 are from earlier analyses. Results for 2006 have been updated from last year’s analysis, when only 40 percent of Form 5500 data were available from the DOL.
Source: Towers Watson.
Measured by asset-weighted median, DB plans again outperformed 401(k) plans in 2007 by about 1 percentage point. Moreover, DB plans substantially outperformed 401(k) plans through the recent bear-to-bull market cycle (2000-2007). Both plan types performed better on the asset-weighted measure than in the plan-weighted analysis (discussed later). Historically, larger plans outperform smaller plans because they have access to a wider variety of investment options and economies of scale and, in the case of DB plans, more investment expertise. Because this measurement focuses on investment results in larger plans, it more accurately tracks the effects of market performance on the average participant.
In Figure 2, we compare rates of return for the largest one-sixth, largest one-half and smallest one-sixth of the plans (based on the amount of assets in the DB plan, to control roughly for the size of the plan sponsor) from 1995 to 2007.
Figure 2. Comparison of plan-weighted median returns by plan size, 1995-2007
Source: Towers Watson.
The 2007 results are similar to those in our past studies. Larger DB plans outperformed smaller plans by roughly 2 percentage points, and large 401(k) plans outperformed smaller ones by 33 basis points. Among large plans, DB plans outperformed 401(k) plans by 46 basis points. Among small plans, 401(k) plans outperformed DB plans by 1.04 percent.
We ran a simple regression to verify the differential effect of plan size on investment returns (see Figure 3). Our results suggest that size has less effect on 401(k) returns than on DB returns. This could be because small DB plans cannot afford as much expertise as bigger plans, while large and small 401(k) plans often offer the same commercially available funds to participants. The association between plan size and investment returns amongst DB plans is strong for all five years of the analysis. While the differences for DC plans were significant in some years, they were of a lesser magnitude than the differences between small and large DB plans. For example, in the 2006 results, this association for both plan types was statistically significant, and for every incremental increase in the log of assets (measured separately for DB and DC plans), annual investment returns increased by 0.54 percent for DB plans but by only 0.16 percent for 401(k) plans.
Figure 3. Simple regression: impact of plan size on investment returns, 2003-2007
Notes: ** and *** indicate significance at 5% and 1% levels, respectively; insignificant otherwise.
Source: Towers Watson.
A look at investment returns with all plans weighted equally
As in previous studies, we also look at investment returns giving all plans equal weight. Returns for 1995 to 2007 are shown in Figure 4.
Figure 4. Plan-weighted median rates of return for DB and 401(k) plans, 1995-2007
* Data for 1995-2005 are from earlier analyses. The 2006 results have been updated since last year’s analysis, when only 40 percent of Form 5500 data were available from the DOL.
Source: Towers Watson.
After stronger performances by DB plans during the 2000-2002 bear market, 401(k) plans outperformed DB plans from 2003 through 2005, as measured by plan-level medians. DB plans and DC plans realized equal returns in 2006, with 401(k) plans taking the lead again in 2007. But over the 13-year period, which captures both bull and bear cycles, DB plans outperformed 401(k)s by an average of 23 basis points.
Over time, 401(k) plans have a wider distribution of returns than DB plans. Figure 5 shows the standard deviation of returns from 1995 to 2007 for both plan types. The standard deviation measures how closely the observations cluster around the mean in a data set. With the exceptions of 2002 and 2003, 401(k) plans had a wider distribution of investment returns in all years in the analysis. This is not surprising — 401(k) plans have millions of participants with varying financial skills choosing different mixes of investments, while DB plans have more consistent investment styles and performance. Participants in 401(k) plans also tend toward extreme asset allocations — all equities or all fixed-income funds, which could also lead to wider swings in returns. 4 Moreover, even if DC accounts were managed optimally, variations would likely persist, as DC plan participants must adapt their investing to their much shorter investment horizon.
Figure 5. Standard deviation of returns across plans, 1995-2007
Source: Towers Watson.
Effect of plan expenses on rates of return
The previous analysis focused on returns based strictly on income performance. DB plans typically report income net of investment expenses. However, expenses for 401(k) plans, including administrative costs, are typically deducted from investment returns. As a result, Form 5500 data do not reflect differences in returns for DB and DC plans arising from embedded non-investment costs in the investment income component. This is especially true for mutual fund investments. In 2008, 38 percent of plan assets in 401(k) plans were invested in mutual funds, compared with only 12 percent in DB plans. 5
Mutual funds for 401(k) plans had an average weighted expense of 66 basis points in 2008. 6 With 38 percent of 401(k) plan assets invested in mutual funds, a reasonable assumption is that these fees reduce rates of return by 25 basis points. According to Towers Watson's 401(k) fee data, roughly one-third of mutual fund fees are actually bundled administrative costs. So 401(k) returns lose an average of 8 basis points due to bundled administrative costs incorporated in investment fees.
Between 1995 and 2007, asset-weighted median returns were 1.07 percent higher in DB plans than in 401(k) plans. To make it an apples-to-apples comparison, we add 8 basis points to 401(k) plan returns for implicit bundled administrative costs. This results in a net difference of almost exactly 1 percentage point.
Percentage of equity may affect rates of return, especially for 2008
On the basis of asset-weighted medians, DB plans consistently outperformed 401(k) plans during the 2003-2007 bull market. But during the 1995-1999 bull market, 401(k) plans outperformed DB plans. This reversal could be due to differences in equity allocations.
As shown in Figure 6, from 1995 through 1999, 401(k) accounts had higher and growing allocations of equity compared with DB plans, so they reaped the rewards of high returns when the market was up. Even so, the difference in returns was less pronounced than one would expect, which might be because DB plan fund managers employed greater diversification or more sophisticated investment techniques to maximize returns.
Figure 6. Equity share and rate-of-return difference for DB vs. 401(k) plans, 1995-2008
Source: Towers Watson tabulations of Form 5500 data, U.S. Board of Governors of the Federal Reserve System, Flow of Funds data (2009) and Investment Company Institute Data (2008).
This advantage to professional investing was particularly striking during the 2000-2002 bear market. Despite having similar or even higher allocations to equities than 401(k) plans, DB plans outperformed 401(k) plans during this period. Many 401(k) plan participants seem to have bought high and sold low in the stock market. Some also might have been heavily invested in company stock, and, when the bubble burst at the beginning of this decade, their returns fell accordingly.
Allocations in equity for DC plans and DB plans seemed to change drastically again for 2007 and 2008, with 401(k) plans again having larger allocations to equity. DB plan allocations to equity seem to have peaked in 2006. DB sponsors started to move away from large equity allocations when equity markets reached their zenith near the end of 2007. During this period, equity allocations in 401(k) plans surpassed those in DB plans, possibly as 401(k) participants failed to rebalance their portfolios. By the end of 2008, we witnessed a stark contrast in allocations to equities between the two types of retirement vehicles. While both realized a decrease in equities due to significant losses in the stock market, DB plan sponsors might have benefited from starting to switch allocations in 2007 and most likely continued the rebalancing of their portfolios during 2008. 7 One would therefore expect DB plans to outperform their 401(k) counterparts in 2008, because DC plans seem to have remained heavily invested in equities relative to DB plans at a time when equity markets realized some of their worst returns in years.
A snapshot of 2008 results
There is considerable interest in how plan investments weathered the recent financial crisis. For calendar-year plans, Form 5500 filings were due Oct. 15, 2009, and the DOL does not release the data until much later. So we adopted a different sampling approach for 2008. We asked sponsors of both a DB and a DC plan to forward us their Form 5500 immediately after filing. (We plan to update the 2008 results with a full sample once the DOL releases its 2008 database, likely next summer.) We obtained data for 79 plan sponsors. In this smaller sample, plan size varies greatly. Assets in the DB plans range from $3.4 million to $40.7 billion, with an average value of $1.4 billion. Assets in the DC plans range from $2.7 million to $22.3 billion, with an average value of $1.1 billion.
Our analysis of plan-weighted medians in this smaller sample found a 93-basis-point advantage for DB plans. 8 DB plans realized median returns of -25.27 percent, while 401(k) plans had median returns -26.20 percent. A distribution of returns is shown in Figure 7.
Figure 7. Distribution of investment returns for DB and 401(k) plans, 2008
Source: Towers Watson.
The distribution for DB returns seems to be wider than that for DC returns at year-end 2008, mostly due to a small percentage of DB plans realizing positive returns. Ten percent of DB plans realized investment returns of -9.99 percent or greater, while none of the DC plans had returns in this range. Roughly 54 percent of DB plans realized investment returns between -20.00 percent and -29.99 percent, while 66 percent of DC plans realized returns in this range. Sixty percent of plan sponsors had better returns in their DB plans than in their 401(k) plans. While it is too early to conclude conclusively that DB plans had better investment performance than 401(k) plans in 2008, it should be noted that during the last bull market (2000-2002), DB plans outperformed 401(k) plans for both the asset- and plan-weighted measures by the largest margin relative to all other years in this study.
Our most recent comparison of investment returns finds that after adding updated data for 2006 and 2007, the overall results remain in line with past studies: DB plans outperformed their 401(k) counterparts by roughly 1 percentage point. In terms of asset-weighted medians, which we feel best represents overall participant experience, DB plans have continuously outperformed 401(k) plans over the recent bull-to-bear cycle (2000-2007).
In terms of plan-weighted medians, 401(k) plans reported the same median returns as DB plans in 2006, and 401(k)s outperformed DB plans in 2007. When looking at plan-weighted returns over the recent bull-to-bear cycle, the results are different from the asset-weighted analysis. In both analyses, DB plans outperformed DC plans during bear markets (2000-2002), but 401(k) plans outperformed DB plans on a plan-weighted basis when the market was booming (2003-2007). Historically, larger plans significantly outperform smaller ones, which could account for DB plans' better showing in a comparison of asset-weighted returns.
In past studies, DB plans outperformed 401(k) plans considerably in bear markets. In the analysis of the smaller subset of plan sponsors for this recent bear market (2008), DB plans seem to be outperforming their 401(k) counterparts again. While our sample is too small for certainty, DB plans’ smaller allocations to equity by year-end 2008 suggest one reason for better investment performance. Participants in 401(k) plans were less likely to rebalance their asset portfolios, leaving them more vulnerable to the recent market declines. Meanwhile, many DB sponsors had already shifted toward more conservative investment strategies, which mitigated their losses, as demonstrated by the few DB plans that realized positive or only very slightly negative investment returns.
DB plan trustees have a fiduciary responsibility for investment performance. They or the professionals they hire usually have considerable financial education, experience and access to sophisticated investment vehicles — advantages 401(k) plan participants typically lack. Plan sponsors and retirement regulators have implemented practices to mitigate the knowledge gap between institutional investors and 401(k) participants. Default funds, typically life cycle/target date funds, are just one example. These new default investment options should moderate the risks 401(k) participants incur by investing in extreme allocations or failing to rebalance their asset portfolios periodically. These regulations are fairly new, and it will take some time to see how these mechanisms affect investment returns and whether they can narrow the gap between DB and 401(k) investment returns.
Even given more investment education and prudent default settings, however, DC plans will never replicate all the advantages of DB plans. At best, these efforts can roughly simulate the investment expertise DB plans enjoy. And DB plans’ longer investment horizon frees them from constraints that are inescapable in a DC plan, such as needing to reduce risk —and its accompanying upside potential — as retirement nears.
1 Earlier analyses appear in "Investment Returns: Defined Benefit Versus 401(k)"
, Insider, June 1998
; "Defined Benefit vs. 401(k) Returns: The Surprising Results"
, Insider, January 2002
; "Defined Benefit vs. 401(k) Returns: An Updated Analysis"
, Insider, September 2003
; "Defined Benefit vs. 401(k): The Returns for 2000-2002"
, Insider, October 2004
; "Defined Benefit vs. 401(k) Plans: Investment Returns for 2003-2006"
, Insider, July 2008
2The 2006 results were based on a partial data set in our prior analysis; here they are based on a complete and final database.
3Comparing investment returns for all DB and 401(k) plans in the DOL Form 5500 database yielded results similar to those presented here.
4“ Investing in Retirement Accounts: Observed Patterns of Asset Allocation,” Insider, August 2009.
5U.S. Board of Governors of the Federal Reserve System, Flow of Funds data (2009).
6Investment Company Institute, “The Economics of Providing 401(k) Plans: Service Fees and Expenses, 2008,” Research Fundamentals, August 2009.
7 See Towers Watson’s “The New Reality of Pension Investment Strategies,” August 2009.
8 Due to the small sample size, results for asset-weighted medians might be misleading because much of the median weight could be attributed to one very large plan. Therefore this measurement is not shown for this subset of employers. We plan to report the asset-weighted median once the analysis is updated with complete data. Using plan-weighted means, DB plans outperformed 401(k) plans by roughly 4 percentage points (-23.29 percent vs. -27.20 percent).