A defined-benefit (DB) pension is the most actuarially-complex product in the typical retirement-planning surface. Unlike a 401(k), a DB pension doesn't have an account balance — it has an accrued benefit, which is a stream of future payments computed by the plan's benefit formula, with a present-value calculation that depends heavily on the discount rate used. Lump-sum offers swing 20-30% across a typical Fed-rate cycle for the same underlying benefit. Joint-and-survivor election decisions are mortality-assumption-sensitive. Early-retirement reductions follow specific actuarial schedules.
This article is the data shape DB-pension-aware test corpora actually need: the benefit-formula inputs (final-pay, career-average, cash-balance), the lump-sum-vs-annuity calculation under the §417(e) segment rates, the joint-and-survivor election framework, and the synthetic households that exercise each path without collapsing to "balance equals X."
What "DB pension" actually means
The benefit formula is plan-specific but typically follows one of three patterns:
| Formula type | How accrued benefit is computed | Common in | |
|---|---|---|---|
| Final average pay | Final-N-years average × accrual rate × years of service | Traditional corporate pensions | |
| Career average pay | Average earnings across all years × accrual rate × years of service | Some corporate plans, especially newer | |
| Cash balance | Hypothetical account balance grown by pay credits and interest credits | Modern hybrid plans |
A typical traditional final-average plan: 1.5% × final-5-year-average pay × years of service. An employee with 30 years of service and a $200K final-5-year average has an accrued benefit of $90,000/year ($7,500/month) starting at normal retirement age.
// Traditional final-average DB pension
{
"plan_id": "DBP-2024-EMPLOYER-001",
"plan_type": "final_average_pay",
"employer": "BigCo Pension Plan",
"employee_id": "E-2010-001",
"vesting_status": "fully_vested",
"vesting_date": "2015-04-12", // 5 years to vest
"service_credit_history": {
"first_service_date": "2010-04-12",
"current_service_credit_years": 14.6,
"service_credit_breaks": [] // periods of inactive service
},
"compensation_history": [
{ "year": 2020, "compensation": 158000.00 },
{ "year": 2021, "compensation": 168000.00 },
{ "year": 2022, "compensation": 182000.00 },
{ "year": 2023, "compensation": 195000.00 },
{ "year": 2024, "compensation": 210000.00 }
],
"benefit_formula": {
"type": "final_average_pay",
"accrual_rate": 0.015, // 1.5% per year of service
"averaging_period_years": 5,
"current_5_year_avg_pay": 182600.00,
"current_accrued_benefit_annual": 39961.20, // 0.015 × 14.6 × $182,600
"normal_retirement_age": 65,
"early_retirement_age": 55,
"early_retirement_reduction_per_year": 0.04, // 4% per year before NRA
},
"joint_and_survivor_options": [
{ "form": "single_life", "factor": 1.0000 },
{ "form": "50_percent_J_and_S", "factor": 0.9100 },
{ "form": "75_percent_J_and_S", "factor": 0.8650 },
{ "form": "100_percent_J_and_S", "factor": 0.8300 }
],
"spouse_age": 58, // for J&S calculation
"lump_sum_offer_available": true
}
The lump-sum calculation and interest-rate sensitivity
When a DB plan offers a lump-sum payment in lieu of an annuity (common at separation, plan termination, or specific window-offer events), the lump sum is computed as the present value of the future benefit stream. The discount rate has dramatic impact:
LS = Σ_t [B × Survival_t / (1 + r)^t]- LS
- = lump-sum offer
- B
- = annual annuity benefit
- Survival_t
- = probability of being alive at age t (per mortality table)
- r
- = discount rate (typically based on Treasury or pension-segment rates)
- t
- = years from offer date
Annual benefit $90,000. 30-year horizon (typical retiree). At 3% discount rate: lump-sum ≈ $1,750,000. At 5% discount rate: lump-sum ≈ $1,380,000. Same benefit, 21% smaller lump sum at the higher rate. The Fed funds rate cycle moves the lump sum dramatically year-over-year.The actual discount rate used is mandated by the plan's terms or by IRS Section 417(e) segment rates for non-cash-balance plans. Section 417(e) uses a three-segment rate structure (segment 1 for 0-5 years, segment 2 for 5-20 years, segment 3 for 20+ years) tied to corporate-bond yields. The segment rates change monthly; a lump-sum offer in March can be 5-10% different from the same offer in June for the same benefit.
For wealth-tech platforms, this means a pension's "value" is not a single number. A DB pension with a $90K/year benefit is worth (in lump-sum terms):
- ~$1.75M if rates are at 2024 levels (~3% segment rate)
- ~$1.40M if rates are at 2008-pre-crisis levels (~5%)
- ~$1.95M if rates are at 2021 lows (~2%)
Test data has to model this rate-sensitivity explicitly. A platform that displays a static "pension value" without flagging the rate dependency produces customer experiences that don't match real-world quotes.
Joint-and-survivor election
For married participants, federal law (REA 1984) requires the default form of payment to be joint-and-survivor (J&S) at 50% or higher unless both spouses formally waive. The participant chooses the J&S percentage; lower percentages get a higher initial benefit but reduced or no benefit to the surviving spouse.
The actuarial-equivalent factors depend on both ages and gender (with sex-distinct mortality tables in non-Sex-Discrimination jurisdictions). For a 65-year-old participant with a 62-year-old spouse, typical 50% J&S factors are around 0.91 — meaning the J&S benefit is 91% of the single-life benefit, with the survivor receiving 50% of the J&S benefit ($90K × 0.91 × 0.50 = $40,950) for life after the participant dies.
The election is irrevocable at retirement — and the right answer depends on:
- Relative ages of participant and spouse
- Health and life-expectancy of each
- Outside resources (life insurance, other retirement income)
- Cost-of-living adjustment provisions (most DB pensions are fixed; some have COLA)
- Estate-planning preferences
A test corpus has to include households across the J&S choice space — single participants (no J&S choice), married participants of various age gaps, married participants with different elections, second-marriage situations with prior-spouse claims under QDRO orders.
Early retirement and the actuarial reduction
Most DB plans allow retirement before the normal retirement age (typically 65 or 67) at a reduced benefit. The reduction varies by plan: a typical schedule reduces the benefit 4-6% per year before NRA, sometimes with breakpoints (e.g., 3% per year between 55 and 60, 6% per year between 50 and 55).
// Early-retirement reduction schedule (example)
{
"normal_retirement_age": 65,
"earliest_retirement_age": 55,
"reduction_schedule": [
{ "age": 64, "factor": 0.96 },
{ "age": 63, "factor": 0.92 },
{ "age": 62, "factor": 0.88 },
{ "age": 61, "factor": 0.84 },
{ "age": 60, "factor": 0.80 },
{ "age": 59, "factor": 0.74 },
{ "age": 58, "factor": 0.68 },
{ "age": 57, "factor": 0.62 },
{ "age": 56, "factor": 0.56 },
{ "age": 55, "factor": 0.50 }
]
}
The customer-facing decision: take a reduced benefit at age 60 ($90K × 0.80 = $72K/year) or wait to age 65 for the full $90K. The break-even calculation depends on assumed mortality and assumed return on the early-payment proceeds (since the customer can invest the early payments through the gap years).
Plan-funding-status modeling
A DB pension's stated benefit can be reduced if the plan is underfunded and goes through PBGC takeover or distress termination. The PBGC guarantees benefits up to a per-month maximum that depends on age (typically $7,000-$8,000 per month at age 65 for plans covered as of 2025). For HNW participants whose DB benefit exceeds the PBGC guarantee, plan funding status matters.
// Plan funding status overlay
{
"plan_funding_ratio": 0.78, // 78% funded as of last valuation
"funded_status_trend": "declining",
"pbgc_guarantee_at_age_65": 88000.00, // PBGC maximum at age 65
"participant_accrued_benefit": 110000.00, // exceeds PBGC max
"pbgc_haircut_risk": 22000.00, // potential reduction if plan terminates
"valuation_date": "2024-12-31"
}
A test corpus that ignores plan-funding status produces over-confident projections for DB participants of underfunded plans — and there are real-world examples (multi-employer plans, several airline plans historically) where the haircut materialized.
What synthetic data has to model
A realistic DB-pension-aware test corpus needs:
| Test case | What it exercises | |
|---|---|---|
| Final-average vs. career-average vs. cash-balance plan | Different benefit formulas; different accrual mechanics | |
| Mid-career service-credit break | Accrual interruption logic; rejoining and prior-service-credit treatment | |
| Lump-sum offer at multiple rate environments | §417(e) segment-rate sensitivity; year-over-year offer variation | |
| Single vs. J&S election decision | Actuarial-equivalent factor application; spousal-protection logic | |
| Early retirement at various ages | Reduction-schedule application; break-even calculation | |
| Underfunded plan with PBGC interaction | Plan-funding-status haircut; benefit-reduction projection | |
| QDRO-divided benefit | Per-spouse benefit allocation; alternate-payee claim handling | |
| Multi-plan participant (multiple DB plans across career) | Cross-plan benefit aggregation; per-plan election independence |
How this shows up in our catalog
The retirement-tagged bundles in the WealthSynth catalog include households with DB pension participation in realistic distribution: ~12M active participants in private-sector single-employer plans per PBGC, concentrated in older cohorts (Boomers and older Gen X) and specific industries (manufacturing, public sector, large legacy corporates). The synthetic plan parameters reflect typical accrual rates, J&S factors, and early-retirement reduction schedules. Plan funding status is included with realistic distribution of funded ratios. The longitudinal data carries year-by-year benefit accrual.
For the broader decumulation context, see Decumulation Edge Cases. For the lump-sum-vs-annuity decision specifically, see Lump sum vs. annuity for pension decisions. For the SERP/NQDC overlay common at senior-executive levels, see NQDC and §409A deferred-comp modeling.