The Center for Retirement Research at Boston College has published a new report comparing investment returns by plan type from 1990 through 2012, using data from the U.S. Department of Labor’s Form 5500.

During the sample period, defined benefit (DB) plans outperformed 401(k)s by an average of 0.7% per year, CRR finds, even after controlling for plan size and asset allocation. As noted by the paper’s authors, much of the money accumulated in 401(k)s is eventually rolled over into individual retirement accounts (IRAs), “which earn even lower returns.”

CRR finds one major reason for the lower returns in 401(k)s and IRAs is higher fees, “which should be a major concern as they can sharply reduce a saver’s nest egg over time.”

Regarding the calculation method, the CRR researchers explain aggregate returns can be calculated in a number of ways. One approach is simply to average the rate of return calculated for each plan in the sample, but this produces a somewhat misleading average because, in reality, a small number of very large plans tend to hold much larger portions of assets than their small-plan counterparts. “So, an alternative measure would weight returns by plan assets and then identify the average,” the paper explains.

Whether the two approaches to calculating returns yield different results depends on the size distribution of plans and the relationship between size and returns. CRR finds a high percentage of plans and participants generally fall into the “<$100 million” category, but the bulk of assets does in fact rest in the largest plans. In the case of DB plans, returns very clearly increase with the size of the plan.

“The pattern is somewhat different for defined contribution plans, where returns increase until plans reach $1 billion and then decline thereafter,” CRR says. “In both cases, excluding plans with less than $100 million will produce higher returns. Weighting by assets will also produce higher returns for both types of plans because it will deemphasize the low returns earned by small plans.”

NEXT: Outperformance expanding? 

CRR researchers note some other recent papers have argued 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.”

Particularly interesting, CRR finds that although the asset allocation of the two types of plans differed significantly over the period 1990 to 2012, asset allocation would still be expected to have only a modest effect on relative returns. “The reason is that the long-run (1926 to 2014) pattern, where risky equities significantly outperformed less risky long-term corporate bonds, has not held over the past two decades,” CRR says.

The researcher explain that, “to account for the differences in allocations to broad asset classes, it is necessary to estimate regression equations in which the dependent variable is the annual return and the explanatory variables include a flag set equal to 1 for a defined benefit plan; a control for the size of the plan; the percentage held in equities; and a variable for each year to account for overall fluctuations in the market.” The results of this analysis show that “both fund size and equity share are associated with higher returns, but, after controlling for these factors, defined benefit plans still earned returns at least 0.7% higher than defined contribution plans.”

CRR says these results hold whether returns are weighted by assets or whether plans with less than $100 million are included or excluded, noting regression equations were also estimated for the sub-periods 1990-2002 and 2003-2012—with the coefficient of the defined benefit variable ranging from 0.3% to 1.5%.

“Overall, the coefficients of the defined benefit flag in the regression equation were very close to those calculated directly from the Form 5500 data,” CRR says. “Thus, neither 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.”

The paper’s authors are Alicia H. Munnell, director of the CRR and the Peter F. Drucker Professor of Management Sciences at Boston College’s Carroll School of Management; Jean-Pierre Aubry, associate director of state and local research at the CRR; and Caroline V. Crawford, a research associate at the CRR. A full copy of the research report is available online here.