Cash Balance Plans: Managing Legacy Liabilities and Looking Ahead with Innovative Designs

• For many plan sponsors, the cash balance (CB) component of their pension liability likely became more material over the last few years as interest rates rose, funded status improved, and/or a portion of traditional defined benefit (DB) liabilities was eliminated via pension risk transfer.
• Under U.S. accounting standards, CB liabilities with bond-based interest crediting rates exhibit very different interest rate sensitivities than traditional DB liabilities and require more complex hedging strategies. 
• Innovative CB plan designs, such as market-return plans, and evolving accounting standards may mitigate, or potentially even avoid, these complexities, potentially making CB plans more attractive to plan sponsors contemplating pension surpluses, benefit enhancements, and retirement income solutions.

Cash Balance Plans: Why Now?

Nearly 40 years ago, Bank of America was the first major plan sponsor to establish a CB DB plan. Since then, thousands of plan sponsors have adopted CB plans, either as a less costly and more predictable alternative to a traditional DB plan or as an attractive, tax-advantaged supplement to a defined contribution (DC) plan. As of 2021, CB plans represent over a third of private pension assets.¹

CB plans are often attractive to both plan sponsors and participants due to their DC-like, account-based structure, transparent interest crediting rates (ICRs), and lump sum form of benefit. Although this account-based view of CB liabilities may seem intuitive, under U.S. funding and accounting regulations, CB liabilities typically also include the present value of future interest credits. This paper addresses key complexities of hedging CB benefits under this approach and focuses specifically on the most common ICRs: fixed-rate and bond-based. Please see Appendix A for background on CB plans.

Despite these complexities, CB plans may not be a relic of the past. Indeed, in the last 12 months, IBM re-opened its CB plan (utilizing traditional ICRs), and three major U.S. airlines incepted brand new CB plans for their pilots (utilizing market-return ICRs). At the end of the paper, we briefly address how market-return CB designs and a supportive regulatory environment may avoid certain challenges of legacy CB designs and help plan sponsors and participants improve retirement security.

Legacy CB Liabilities: Interest Rate and Credit Spread Sensitivities

As we noted above, plan sponsors often think of their CB liability as the sum of the hypothetical participant account balances. From that perspective, the investment objective may simply be to outperform the ICR with the least amount of risk without any specific liability to hedge. Many small businesses and professional services firms, who are usually more focused on maximizing tax-deductible contributions rather than managing balance sheet volatility, do precisely that.

However, under U.S. accounting standards, the present value of CB benefits includes not only the account balance but also all assumed  future interest credits until the participants’ expected retirement dates, discounted to the valuation date. This accounting liability may be larger or smaller than the sum of the hypothetical account balances. As discussed in Appendix A and summarized in Figure 1, this liability methodology also results in unique and unintuitive interest rate sensitivities. However, in many cases these sensitivities can be hedged:

• Fixed-rate: CB liabilities with fixed-rate ICRs are straight-forward, as their interest rae sensitivites are like those of traditional DB liabilities. 
• Bond-based: Hedging these ICRs is more complex, but certainly achievable, and we devote the bulk of this paper to these ICRs.
• Floors: As floors create optionality, close hedging is more complex, costly, and beyond the scope of this paper.

Figure 1. Sensitivities of CB Liabilites (Assuming 100% Lump Sum Election)²

Source: Dodge & Cox

Hedging Bond-Based ICRs

Consider a hypothetical $1 billion, fully funded, standalone CB plan with the actual and assumed future ICRs equal to the 10-year Treasury yield (highlighted row in Figure 1).⁴ The liability-hedging assets (LHA) for this plan could consist of two components:

• An investment-grade (IG) credit portfolio that achieves the target credit spread hedge ratio, and
• A Treasury-derivative overlay to fully hedge out the credit portfolio’s Treasury duration.

The unusual key rate duration profile of CB liabilities, specifically large negative duration at the 10-year tenor (see Figure 2), requires a nuanced approach to the derivatives overlay.

Figure 2. Sample Plan Sensitivities

Source: Dodge & Cox. KRD is key rate duration.

Assuming no return-seeking assets (RSA) for illustrative purposes, Figure 3 shows a potential liability-hedging portfolio.

A few observations:

• Achieving a near-zero duration requires the custom Treasury portfolio (or a completion portfolio) to have a duration of –39 years. This would typically be achieved via a Treasury derivatives overlay that is supported by cash and/or physical Treasuries.
• Assuming a credit beta of 1.33 between IG and AA spreads, an 85% allocation to a U.S. Credit mandate (benchmarked to the Bloomberg U.S. Credit Index⁵) could achieve a credit spread hedge ratio (CSHR) of 92%. A bespoke blend of Long Credit and Intermediate Credit, or inclusion of higher-beta fixed income assets, such as below investment-grade credit, could achieve a CSHR close to 100%.
• The allocation between credit, Treasuries, and Treasury derivatives could vary depending on the plan sponsor’s target hedge ratios, comfort with leverage, and liquidity needs, among other considerations.

A more realistic asset allocation would likely include RSA. By accounting for the correlation between RSA returns and changes in AA credit spreads in the CSHR, a shorter-duration credit portfolio (e.g., Intermediate Credit) may also make sense. In turn, this should result in needing to hedge out less duration with derivatives, reducing the implied leverage in the custom Treasury portfolio.

The prospect of a Treasury overlay with a large negative duration may be daunting at first. However, since CB liabilities are often part of a larger DB plan or a master trust with traditional DB plans, the positive duration and key rate durations of these DB liabilities offset the complexities of CB liabilities at the total plan or master trust level. In addition, even standalone CB plans may include retiree liabilities if some participants elect an annuity form of payment. 

Figure 3. Sample Plan Liability-Hedging Asset Allocation⁶

Source: Dodge & Cox

Figure 4 summarizes liability sensitivities for a sample $1 billion fully funded plan consisting of 20% CB liabilities as described above and 80% traditional DB liabilities, while Figure 5 shows a potential asset allocation, assuming a more typical 80%/20% LHA/RSA split. One caveat: many plans define the benefit at retirement as the larger of a CB benefit and a frozen DB benefit. In that case, it may be impossible to “separate” CB and DB benefits this cleanly, and interest rate sensitivities may need to be estimated at the total plan level. The CB benefit at retirement grows at the assumed future ICRs and may be larger or smaller than the fixed, frozen DB benefit depending on the assumed future ICRs.

Figure 4. Sample Plan Sensitivities

Source: Dodge & Cox. KRD is key rate duration.

Figure 5. Sample Plan Liability-Hedging Asset Allocation

Source: Dodge & Cox

A high-level conclusion of this analysis is that if the plan sponsor sought a 100% interest rate hedge ratio without accounting for the zero-duration nature of the CB liabilities, the plan would likely be significantly overhedged–by more than 20%. This explains why in the current environment of relatively high interest rates and strong funded status, it is critical to accurately account for CB liabilities.

Ongoing Monitoring

Asset-liability monitoring is a key component of prudent liability hedging. At a given point in time, traditional DB liability characteristics can be estimated by (1) discounting liability cash flows produced at valuation date with current market interest rates and (2) accounting for benefit payments and accruals between valuation date and analysis date. The same approach applies to CB plans with fixed-rate ICRs, as the associated liability cash flows are independent of changes in interest rates.

However, for bond-based ICRs, future assumed ICRs and liability cash flows vary with market conditions. In this case, discounting cash flows produced at the time of valuation may not accurately reflect liability present value and characteristics at a given point in time, especially if interest rates have changed significantly since the valuation date.

On the other hand, generating full-valuation liability cash flows on an ongoing (say, monthly) basis may also not be practical. Instead, it may be sufficient to generate liability cash flows at the time of valuation under several interest rate assumptions and, as market conditions evolve, interpolate between them. This approach also accounts for “larger of” benefit formulas mentioned in the previous section. While there is certainly a loss of precision compared to a full valuation approach, it may be small relative to unhedgeable characteristics, like plan experience and demographic assumption changes.

Reducing CB Liabilities via Pension Risk Transfer and Lump Sum Windows

Given the complexities associated with hedging CB liabilities with bond-based ICRs, plan sponsors may wish to “off-load” some or all of these liabilities via a pension risk transfer (PRT). However, precisely because of these complexities and the uncertainties associated with participant behavior (e.g., when to take the benefit and whether in lump sum form or converted to an annuity—this applies to fixed-rate ICRs as well), PRT for CB liabilities is typically much more expensive (relative to the accounting liabilities) than it is for traditional DB liabilities. 

Offering a lump sum window may be a more effective way to reduce CB liabilities, but it is usually also costly relative to the accounting liabilities. The lump sum is typically equal to the account balance, while the accounting liability reflects the “project forward and discount back” methodology. Thus, since bond-based ICRs typically reflect Treasury yields while liability discount rates reflect AA corporate bond yields, the account balance is usually smaller. (Depending on the level of corporate spreads, this may not be the case for bond-based ICRs that reflect a spread to Treasuries.) The same logic applies to fixed-rate ICRs: in the current environment where liability discount rates are hovering around 5%, it may be advantageous to offer lump sums for ICRs greater than 5% and disadvantageous otherwise.

Looking Ahead: CB Designs

Many of the challenges associated with managing CB liabilities would disappear if (1) the ICR were equal to actual asset returns and (2) the accounting (and funding) liability were equal to the sum of hypothetical account balances.

On the first item, IRS regulations already permit market-return ICRs, subject to diversification (in accordance with ERISA) and the preservation of capital rule (in accordance with the Pension Protection Act). Under these rules, the benefit payable to the participant at retirement may not be less than the sum of the pay (or service) credits. This is close, but not equivalent, to a cumulative zero-percent return floor. “Market-return” CB plans are fairly common among employee-owned professional services firms.

The second item remains unresolved, but there have been encouraging developments recently. In February, the Conference of Consulting Actuaries—an influential industry group that includes many actuaries affiliated with large accounting firms—issued a position paper that advocates for account-based treatment of market-return CB plans for accounting purposes.⁷ In addition, the IRS recently released guidance⁸ that addresses certain regulatory aspects of market-return CB designs that could pave the way toward “account-based” treatment from a funding perspective as well.

Should these issues be fully resolved, we believe plan sponsors might find CB plans attractive for several business reasons: 

• Enhancing employee benefits. By tailoring the asset allocation and pay- (or service-) based credits to specific plan sponsor objectives and participant population characteristics, the plan could serve either as a baseline, low-risk retirement plan or a substantial supplement to a DC plan. It could also aid in attracting and retaining talent.
• Offering a plan that automatically includes an annuity option, avoiding the challenges of offering a retirement income solution in a DC plan. Since a CB plan is subject to DB regulations, participants can elect to take their lump sum as an annuity.⁹
• Allowing more flexible plan sponsor funding of retirement savings compared to a DC plan. Since a CB plan is subject to DB funding rules, plan sponsors could pre-fund several years of pay- (or service-) based credits in profitable years, and in unprofitable periods, they could rely on that pre-funding (i.e., plan surplus) or reduce contributions.

There are considerations, including the preservation of capital rule, vesting requirements, PBGC¹⁰ premiums, and population dynamics.

IBM’s re-opening of its surplus-heavy CB plan and the attendant shift of DC contributions to CB contributions and the airlines’ inception of market-return CB plans illustrate practical applications of these objectives. IBM’s crediting rate is 6% for the first 3 years and the 10-year Treasury yield with a 3% floor after that. In the case of Delta Airlines, the initial asset allocation reflects 60% fixed income/40% equities. We encourage plan sponsors to inquire with their actuary about the evolving CB landscape and evaluate whether these developments may be of benefit to them.

Conclusion

CB plans may be in the spotlight, either as a complex legacy liability or as an innovative opportunity to enhance retirement security. The former often exhibits unusual interest rate sensitivities, such as near-zero duration (but not key rate durations!), and requires a hedging strategy that is quite different from that of traditional DB liabilities. The latter avoids these complexities by aligning benefits with the plan’s asset returns and, in an evolving but generally supportive regulatory environment, may be worth investigating further.

We would welcome the opportunity to speak with you or your advisor about our pension risk management solutions as you proceed on your pension journey. 

Appendix A: Legacy Cash Balance Liability Sensitivities

Cash Balance Background

In a CB plan, the plan sponsor establishes a hypothetical account for each participant and credits that account with annual pay (or service) credits and interest credits. The normal form of benefit at retirement is a lump sum equal to the hypothetical account balance, although the participant may elect to receive their benefit as an annuity. Crucially, if the plan is frozen and/or if the participant terminates employment but remains in the plan, interest rate credits continue, while pay (or service) credits cease.

Under IRS regulations, permissible ICRs include fixed rates; variable rates based on Treasury bond yields, IRS funding segment rates, and the Consumer Price Index¹¹; and variable rates based on market returns, subject to certain conditions, including a preservation of capital rule. These rules provide that the lump sum payable at retirement may not be less than the sum of the annual pay (or service) credits. In addition, certain spreads to the variable rates as well as floors and ceilings are also permitted. In our experience, most large CB plans reflect fixed rates or rates based on Treasury bond yields, with or without a floor.

Liability Valuation

Under U.S. accounting standards and funding requirements, the CB plan liability includes the current hypothetical account balances and the present value of all assumed future interest rate credits until each participant is expected to retire, subject to mortality, termination, retirement, and discount rate assumptions. 

This leads to a “project forward and discount back” approach, where the participant’s hypothetical account balance at valuation is projected forward to retirement with assumed (or, in the case of fixed rates, known) future interest credits and discounted back to present value using the (known) accounting discount curve. For example, in a frozen plan with a single participant expected to retire at time T with 100% certainty, the present value of the participant’s liability is

where ICR_T is the assumed interest crediting rate in year t and DR_T is the T-year discount rate at the time of valuation. Thus, like traditional DB liabilities, CB liabilities are sensitive to changes in the discount curve, but unlike traditional DB liabilities, CB liabilities are also sensitive to changes in the assumed ICRs.

Fixed ICR

With a fixed ICR, the future value of the CB benefit at retirement is independent of Treasury yields or credit spreads, and all assumed future ICRs are equal to the fixed-rate ICR in the plan document. Consequently, the present value is sensitive only to changes in the discount rate, and its duration and spread duration are like those of traditional liabilities, albeit somewhat shorter given the lump sum nature of the benefit. This is by far the easiest CB liability to hedge.

Treasury-based ICRs with No Floors or Ceilings

Historically, for the “project forward and discount back” calculation, assumed ICRs beyond the first year were often set equal to a long-term average of the applicable Treasury yields. While this approach is simple and smooths out variability if Treasury yields remain range-bound, it can result in significant year-to-year balance sheet gains and losses if actual interest rates deviate from long-term averages or if the long-term average assumption needs to be changed. This was particularly evident in the 2020-2023 period, when the 10-year Treasury yield rose nearly 300 basis points.

To avoid these complications, many pension practitioners have moved to setting the assumed future ICRs equal to the actual (known) first-year ICR and updating this assumption as the underlying Treasury yields evolve. While nuances still exist, this approach forecasts future benefit payments based on the current market environment and results in a more hedgeable liability.¹² In this case,

• Liability duration is essentially zero since a parallel shift in the Treasury yield curve affects the entire discount curve and the ICR, and the two effects essentially cancel each other out. (In the above equation, the numerator and denominator are affected similarly.)
• Spread duration is positive and large since a Treasury-based ICR is independent of changes in credit spreads, and a change in credit spreads affects only the discount curve. (In the above equation, only the denominator is affected.)
• Key rate durations are not zero and depend on the time series distribution of the cash flows and the tenor of the ICR. The key rate at the tenor of the ICR is typically negative, while the other key rates are positive. 

Figure 6 shows an indicative key rate duration profile when the ICR is equal to the 10-year Treasury yield. Intuitively,

• If the 10-year Treasury yield rises in isolation, when projecting forward, all participants’ benefits increase (as they are projected forward at the higher ICR), but, when discounting back, this increase is offset by a higher discount rate only for benefits due 10 years from now. Consequently, the total liability is higher.
• If any Treasury key rate other than the 10-year rises in isolation, the future value of the liability does not change (since the ICR does not change), but the discount rate at the tenor in question increases, lowering the present value of liabilities due that year and resulting in a lower liability overall.

Figure 6. Sample Cash Balance Plan Key Rate and Spread Key Rate Durations

Source: Dodge & Cox

Setting future assumed future ICR’s based on the forward curve (rather than equal to the first-year ICR) can provide an even more precise estimate of future benefit payments. Continuing with the example above, in an upward-sloping curve environment, the 10-year Treasury yield credited in years two and later is likely to be higher than the 10-year Treasury yield credited in year one. Thus, assuming the actual curve evolves according to forwards, not incorporating the forward curve would underestimate the liability and result in an actuarial loss in year two. Under this approach, top-level risks—duration and spread duration—remain similar to the approach described above, but key rate durations may differ.

This methodology leads to additional calculation complexity and requires even closer coordination between the actuary, accountant, and investment advisor, potentially outweighing the benefits of its greater precision. Further, the magnitude of this enhancement could be modest if the Treasury curve is relatively flat beyond the tenor of the ICR and could be overshadowed by larger interest rate movements as the forward curve is often not an effective predictor of future interest rates.

Treasury-based ICRs with Floors and Ceilings

Complications arise if the ICR also includes a floor or a ceiling. Since under U.S. accounting standards, liabilities are calculated under a deterministic approach, there are essentially two regimes with a “regime change” at the floor:

• When Treasury yields are below the floor, a fixed-rate future ICR assumption equal to the floor is reasonable, and liability sensitivities reflect those of a fixed-rate ICR.
• Otherwise, the approach described above is reasonable, and liability sensitivities reflect those of a Treasury-based ICR without the floor.

A similar situation arises with a ceiling.

A stochastic or options-pricing-based approach to forecasting interest rates, and therefore the ICR, would lead to a more accurate estimate of the liability present value and could lead to an options-based LH strategy that avoids switching between the two regimes. However, this approach is complex, does not align with the deterministic accounting assumptions commonly used in the industry, and would be difficult to implement on an ongoing basis. 

Other ICRs

Other ICRs, including the IRS Segment Rates and the Consumer Price Index, are beyond the scope of this Appendix. As with Treasury-based ICRs, we encourage plan sponsors to collaborate closely with their actuary, accountant, and investment advisor(s) to select future ICR assumptions that reflect underlying market conditions, minimize actuarial gains and losses, and enable alignment with the LH strategy. For the IRS Segment Rates, for example, this would likely mean incorporating an assumption not only about Treasury yields but also credit spreads, which would in turn affect the spread duration of the liabilities.

Appendix B

Overall Disclosure. For illustrative purposes only. The hypothetical information presented does not represent actual results of any client and is based upon the hypothetical assumptions described below. While we believe the assumptions and methods used in this analysis are reasonable, other assumptions may also be reasonable and may lead to results that differ significantly from those shown here. Assumptions have been made for modeling purposes and are unlikely to be realized; not all assumptions have been stated or fully considered; and changes in assumptions may have a material impact on the hypothetical results presented. Actual results for investors will differ from the results contained in this analysis. No investment strategy or risk management technique can guarantee returns or eliminate risk in any market environment.

Sample Standalone Cash Balance Plan Characteristics: Present value of liabilities: $1 billion. Discount rate: 4.95%, FTSE Above Median AA, as of December 31, 2023. Interest crediting rate: 4.50%, 10-year constant maturity Treasury as of November 30, 2023. Plan duration: -0.04, plan spread duration: 8.57. The sample plan is frozen. Calculations do not account for differences between par and spot curves. Dodge & Cox calculated the characteristics of the liabilities. Sources: FRED, FTSE Russell, Dodge & Cox.

Sample Combined Plan Characteristics: Present value of liabilities: $1 billion, 20% in standalone cash balance component with characteristics and assumptions as above and 80% in standalone traditional defined benefit component. DB plan discount rate: 4.99%, FTSE Above Median AA, as of December 31, 2023. Plan duration and spread duration: 10.35 years. The sample plan is frozen. Dodge & Cox calculated the characteristics of the liabilities. Sources: FTSE Russell, Dodge & Cox.

Hypothetical Portfolios: Assumed liability-hedging assets consist of U.S. Credit, Long Credit, and/or Intermediate Credit represented by the Bloomberg U.S. Long Credit Index,¹³ the Bloomberg U.S. Credit Index, and the Bloomberg U.S. Intermediate Credit Index,¹⁴ as shown, and a hypothetical custom Treasury portfolio. The custom Treasury portfolio consists of U.S. Treasury bonds, Treasury derivatives (e.g., futures), and cash. Credit beta of investment-grade credit to the liability discount rate is 1.33.