Retirement Wealth in Defined Benefit and Defined Contribution Pension Plans

10/20/2009
Featured in print Bulletin on Aging & Health

It is often said that a comfortable retirement rests on the "three-legged stool" of Social Security, employer-provided pensions, and personal savings. Significant changes to any of these three components may either enhance or threaten the fi-nancial security of retired workers.

Over the past two decades, there has been a striking shift in employer-provided pensions from defined benefit (DB) to defined contribution (DC) plans. In a DB plan, a worker's benefit is determined by his work history and the specifics of his employer's plan, while in a DC plan, a worker's benefit is determined by his own and his employer's contributions and by asset returns.

The shift from DB to DC plans may affect not only the average value of retirement wealth but also the distribution of possible outcomes, since DB and DC plans are subject to different types of risk. In DC plans, workers face financial market risk, while in DB plans, workers face the risk of unexpected shocks to earnings, job changes, and early retirement. Without careful study, it is difficult to gauge the relative risk of these two types of plans.

In "Defined Contribution Plans, Defined Benefit Plans, and the Accumulation of Retirement Wealth" (NBER Working Paper 12597), authors James Poterba, Joshua Rauh, Steven Venti, and David Wise take up this question. The authors simulate both the average level of retirement wealth and its distribution under DB and DC plans. They do so using actual pension plans that cover respondents in the Health and Retirement Study (HRS) and the actual earnings histories of these individuals, as well as historical data on asset returns. This enables them to incorporate both financial market risk and earnings history risk into their analysis and to compare their relative importance.

The authors begin with a sample of married couples from the HRS. To simulate DC plan balances, each worker is first randomly assigned a contribution rate based on the observed distribution of rates in the HRS. The typical combined employer and employee contribution is 7.7 percent of salary, but the rate ranges from 3 percent at the 10th percentile to 15 percent at the 90th percentile. Next, workers are assumed to invest in three assets - corporate stock, nominal long-term government bonds, and inflation-indexed long-term bonds (TIPS); simulations are conducted for seven portfolios involving different combinations of these assets. Finally, asset returns for each year are drawn from the empirical return distribution. To obtain a distribution of wealth values, DC balances are simulated 50,000 times for each worker in the sample.

The authors then turn to their simulation of DB plan balances. The authors use the earnings histories and detailed pension plan descriptions collected by the HRS to calculate the stream of future DB pension benefits for each worker, assuming that he is assigned to a randomly selected DB plan for each of his jobs. Using information on mortality rates and discount factors, these future benefits are converted into an expected present discounted value of DB pension wealth.

The authors have several key findings. First, mean wealth in DC plans is somewhat higher than mean wealth in private sector DB plans - $177,000 vs. $156,000 for a worker with a high school education - even when DC balances are invested conservatively in an all-TIPS portfolio. When DC balances are invested in corporate stock, the disparities are far greater - the mean wealth in an all-stock portfolio is $919,000. Public sector DB plans compare somewhat more favorably to DC plans, as the mean wealth for a high school graduate under these plans is $327,000.

Comparing the full distribution of outcomes under the two plans, DC balances exceed private sector DB balances except at the very low end of the distribution. However, when the authors reduce the rate of return on corporate stock by 300 basis points to allow for the possibility that future returns may be lower than past returns, DB and DC wealth values are much closer.

While the risk of obtaining a lower wealth level in a DC plan is relatively small, workers might still prefer the DB plan if they are sufficiently risk averse. To get at this question, the authors compute certainty equivalents, which measure the amount of certain wealth that would make the worker as well off as being in the DB or DC plan with all the associated risks. They find that the certainty equivalent for DC plans greatly exceeds that for DB plans for plausible levels of risk aversion. The only case where DB plans have a higher certainty equivalent is for a risk-averse individual in the public sector when future equity returns are lower than past returns.

The authors note that there are a number of simplifications in their analysis. For example, in their analysis workers never withdraw DC balances as a lump sum distribution when changing jobs, though in reality many workers do this, reducing their DC wealth at retirement. Further, their analysis does not allow the length of the worker's career or the worker's wages to depend on his pension plan arrangements. They note that further insights on these issues "are essential for understanding the long-run effect of the ongoing transition from DB to DC pension plans."


The authors gratefully acknowledge funding from the National Institute on Aging (Grant P01 AG005842).