Understanding the Substitution Effect

The preceding paper in our Making the Benefit Connection series explored how suitable glide paths might differ between two otherwise identical DC plans when participants have varying levels of access to a sponsor’s DB plan. ¹In our simulations, we found that suitable glide paths for participants with access to both DB and DC plans typically had lower equity levels throughout the entire investment life cycle because of what we call the wealth effect—the reality that as retirees become better funded (wealthier), they have less need to expose themselves to riskier, more volatile assets in hopes of earning higher returns.

Compared with counterparts who only have access to a sponsor’s DC plan, participants who also have DB plan coverage should tend to be more amply funded for retirement, thanks to the value of their DB plan benefits. Figure 1 illustrates the potential impact of this wealth effect by comparing the optimal glide paths in our simulations for a hypothetical standalone DC plan and for the same DC plan when participants also were covered by a hypothetical companion DB plan.

The hypothetical DC plan shown in Figure 1 was assumed to be a safe harbor design featuring an employer match of up to 100% of the first 3% of salary in employee deferrals and 50% of the next 2%. The hypothetical companion DB plan offered a retirement benefit equal to 1% of the final five‑year average salary per year of service.²

In our simulations, we found that because participants with access to the hypothetical DB plan were better prepared financially for retirement, their DC target date glide path could maintain up to 23 percentage points less equity exposure and still give them a reasonably strong potential of meeting their retirement spending goals. 

While the results in Figure 1 are interesting on their own, the hypothetical comparison used in our simulations admittedly was somewhat unrealistic. It assumed that the DB plan offering was strictly supplemental to the sponsor’s DC plan rather than a substitute for a more generous DC plan. 

However, plan sponsors designing retirement programs to serve as recruitment and retention tools ultimately are constrained in their design choices. Any sponsor will have a finite budget for retirement benefits, which raises numerous questions and implies potential trade‑offs.

• How can the benefit budget best be deployed to align with the organization’s retirement philosophy? 
• Does the sponsor want to encourage employees to retire in their late 50s, their early/mid‑60s, or later? 
• Should retirement benefits be tied to the success of the organization? 
• How does the sponsor weigh the relative importance of retirement outcome predictability versus cost predictability? 

Given these constraints, we believe that a realistic assessment of glide path suitability requires that a DB plan should not be evaluated simply as an additional benefit paired with an existing DC plan, but rather should be compared with an equivalent‑cost DC plan in isolation. 

Identifying equivalent‑cost DB and DC plans

Comparing benefit costs within a single plan type with the same structure can be relatively straightforward. For example, we can definitively say that a defined benefit plan that provides a retirement benefit equal to 1% of final five‑year average salary for each year of service is less generous than the same plan but with a 1.5% salary multiplier. Similarly, a DC plan with a 5%‑of‑salary non‑elective employer contribution is more expensive than a plan with a 3%‑of‑salary contribution. 

However, cost equivalency becomes slightly more difficult to assess when comparing different designs within the same type of plan:

• A 1% final average pay plan, for example, is likely to cost more than a 1% career average pay plan because salaries for most employees will tend to increase throughout their working careers. 
• Similarly, a DC plan that matches participant contributions dollar for dollar up to 6% of salary is likely to cost more than a DC plan that simply makes a non‑elective contribution equal to 2% of salary—although that might not be true if many employees don’t participate in the plan.

Determining cost equivalency across plan types (in this case, between DB and DC plans) adds another level of complexity to the exercise, requiring a myriad of assumptions—including, but not limited to, expected investment returns, interest rates, employee participation rates, deferral elections, salary growth rates, actuarial funding methods, participant mortality, estimated retirement ages, termination incidence, disability incidence, and more.

Additionally, funding mechanisms differ for U.S.‑based DB plans based on whether the plan is sponsored by a corporate or a government entity.³ These funding decisions are an additional wrinkle that needs to be considered in any cost comparisons. 

Different funding methods mean that sponsor contributions to a DB plan trust will be made at different points in the employment cycle, giving those investments differing amounts of time to generate the returns needed to pay future benefits. ⁴ This nuance is subtle, but it is why there are multiple “equivalent cost” lines for the hypothetical DB plans shown in Figure 2, depending on how the plan sponsor was assumed to fund those benefits. 

Based on insights gleaned from the aggregate behavior of the 2.2 million participants in T. Rowe Price’s recordkeeping database, our simulations indicated that the cumulative 40‑year cost of a hypothetical 1% final average pay DB plan (without service limit) based on corporate funding rules was approximately equal to the cost of a DC plan with a 10.7% salary match/non‑elective deferral. The 40‑year cost of an equivalent hypothetical DB plan based on public funding rules, meanwhile, was approximately equal to a DC plan with a 9.4% salary match/non‑elective deferral. 

Glide path results

In our next set of simulations, we sought to control for potential differences in benefit richness (i.e., for the substitution effect⁵) by enhancing our baseline hypothetical DC plan. The enhanced plans offered the original matching formula but also included the additional nondiscretionary contributions (10.7% or 9.4%) required to equal the cost of our hypothetical DB plan under both corporate and public funding rules.

In our simulations, these changes eliminated about half of the difference in optimal equity exposure between the baseline DC‑only glide path and the DC‑plus‑DB glide path (Figure 3).

Stated differently, the wealth effect seemed to explain about half of the difference in glide path equity exposure caused by the existence of a DB plan. In fact, a 20% nondiscretionary contribution was required in our simulations to generate a hypothetical DC‑only glide path with equity levels comparable to our combination of a baseline/safe harbor DC plan and the final average pay DB plan. 

The other half of the difference in glide path equity was explained by the benefit accrual and payment structures themselves. A DC plan with nondiscretionary contributions provides extra savings during working years and then becomes the primary source of income for most participants after retirement (even though Social Security benefits typically replace a higher share of earnings for lower‑paid employees). A DB plan provides a relatively stable and secure source of retirement income that shields participants from market volatility. This should tend to reduce their reliance on their DC plans and other savings.

Under these circumstances, when the DC plan in our simulations provided benefits comparable in employer cost to the final average pay DB plan (in other words, fully reflecting the substitution effect), the optimal equity level in the DC‑only glide path still was higher than in a glide path designed for participants who also had access to the company’s DB plan. 

By contrast, even when total benefit costs were comparable, providing part of that benefit in the form of a DB plan resulted in lower equity levels in the DC target date glide path in our simulations, particularly at the time of retirement. The DB plan benefit structure helped maintain wealth stability, which itself has potential utility value for participants. 

Conclusions

In isolation, adding a hypothetical DB plan to an existing hypothetical DC plan substantially reduced the optimal equity allocation in the target date glide path in our simulations. However, looking at the DB plan in isolation overly simplifies the trade‑offs that many plan sponsors actually face. 

DB plan closures and freezes are continuing and not many new plans are being offered, particularly by larger employers. Instead, many plan sponsors are enhancing their DC plans, either by improving percent‑of‑salary matching rates or by increasing non‑elective contributions to offset the end of DB coverage.

To us, these trends indicate that DB plan coverage doesn’t necessarily make participants better funded relative to those who only have access to a DC plan, given that there is an implicit trade‑off at play. When we controlled for overall benefit costs in our simulations, we found that the impact of the DB plan on the optimal DC plan glide path still was to reduce equity exposure, but not by as much as if we had ignored the substitution effect entirely. 

The substitution effect explains about half of the difference in glide path equity allocations in our simulations, with the remaining half a result of structural differences in benefit designs between the DB and DC plans that we analyzed. We think this finding could be useful for plan sponsors who want to look holistically at their retirement benefit structures and, more importantly, consider the impact of their DB plan on their DC plan glide path.