Evidence

Methodology

Scope

The 2025 CPS Annual Social and Economic Supplement reports income received during calendar year 2024. Wage excess is calculated person by person as the released wage value minus $168,600, floored at zero, then aggregated with the official person weights.

Birth cohorts and uncertainty

The point birth year is 2025 minus age at survey. The true birth year can be one year earlier depending on whether the respondent had a birthday before the March survey. The dashboard groups those estimates into five-year cohorts to reduce distracting volatility from sparse single-year cells. Standard errors use all 160 CPS ASEC replicate weights and the Census successive-difference variance factor, with 90% confidence intervals.

Confidence labels

The table’s confidence label is a plain-language reliability guide for each wage-excess interval, based on the number of unweighted above-cap CPS records and the relative standard error implied by the published interval. N/A means no above-cap sample cases. Low means fewer than 30 cases or relative standard error above 30%. Medium means at least 30 cases with relative standard error no greater than 30%. High is reserved for at least 100 cases with relative standard error no greater than 15%. These labels describe statistical precision, not certainty that the estimate is correct.

One-year pilot proposal

The proposal calculator models a fixed pilot cohort of 2024 wage-and-salary workers with earnings at or above the taxable maximum. Self-employment income and SSA’s 22.119 million self-employed-worker count are excluded. The model applies one selected OASDI contribution rate to one year of wage excess, allocates that contribution among cash, bonds, and equities according to the selected bucket years, and distributes the fixed cohort’s refundable contributions across retirement years using CPS birth-year shares. It does not yet model a recurring annual program, future worker cohorts, population growth, wage growth, mortality, retirement behavior, taxes, fees, benefit credits, or market volatility.

The default 6.2% rate applies only the employee OASDI share to wage-and-salary earnings above the taxable maximum. The pilot does not impose a matching assessment on employers and does not include self-employment income. It therefore models a new employee contribution only for wage earners above the maximum, while preserving the current employer and self-employed treatment for this illustration. The editable rate can still be changed for comparison scenarios.

Purchasing-power convention

All calculator cash flows, balances, and return assumptions are expressed in constant 2024 purchasing-power dollars. The repayment promise is therefore 100% of each modeled contribution adjusted for inflation, using the same CPI-W convention that Social Security uses for annual cost-of-living adjustments. In future nominal dollars, the amount returned would rise with the intervening COLAs; in this constant-dollar ledger, that inflation adjustment is already reflected by keeping the real contribution amount unchanged.

The default 2.5%, 4.0%, and 7.5% investment assumptions are real returns after inflation. Inflation indexing preserves a worker’s purchasing power but provides a zero-percent real return on the refundable contribution itself. It does not compensate the worker for lost liquidity or the investment earnings the worker might otherwise have received. Investment earnings above the inflation-adjusted worker obligation remain in the reserve.

Ten-year refund schedule and bond ladder

The default scenario divides each worker’s inflation-adjusted contribution into ten equal annual installments beginning at the assumed retirement age. In constant 2024 dollars, each installment is one-tenth of the refundable contribution; its future nominal payment would reflect the applicable CPI-W adjustments. The comparison setting returns the entire amount as a lump sum at retirement.

The ten-year payment period intentionally aligns with the recommended ten years of bonds in the default retirement allocation. This is a liability-matching convention: a ladder of bond maturities can support scheduled refunds over the coming decade, cash can cover the nearest payments, and equities can remain invested against longer-dated obligations. The bucket-year inputs are normalized into portfolio weights, so the calculator illustrates that structure rather than constructing individual bonds or modeling duration, interest-rate risk, or reinvestment risk.

Installments reduce annual liquidity demands and leave more assets invested for longer, but that benefit to the reserve comes from delaying worker access. The model assumes any unpaid refundable balance remains an obligation to the worker or the worker’s estate; mortality does not erase it.

The default 30-year reserve allocation uses 5 years of cash, 10 years of bonds, and 15 years of equities. Normalizing those values produces allocations of 16.67%, 33.33%, and 50%. At assumed returns of 2.5%, 4.0%, and 7.5%, the initial weighted return assumption is 5.50%. The 30-year total is a conservative planning horizon for a government program, not a modeled withdrawal rule.

Passive-investment mandate

The proposal assumes that both bond and equity exposure is implemented exclusively through low-cost, total-market index funds. The mandate is intended to minimize fees, security-selection discretion, industry favoritism, and market timing. It does not authorize individual-stock selection, sector tilts, or policy-driven exclusions.

Passive indexing reduces but does not eliminate governance risk. A public reserve would still require independent oversight, transparent benchmarks and fees, restrictions on political direction, rules for proxy voting and stewardship, and statutory protection against diversion of assets. The calculator does not model those institutional arrangements.

Annual cash-flow sequence

The annual ledger begins with the prior year’s closing bucket balances. It adds the one-time contribution in 2024 only, pays that year’s scheduled inflation-adjusted installments from cash first and then bonds and equities if required, applies each bucket’s fixed real annual return to its post-payout balance, and rebalances the closing reserve to the target allocation. Returns compound annually because each year begins with the preceding closing balance. Rebalancing is assumed to have no transaction costs or tax consequences.

Fixed real annual returns are a simplified deterministic demonstration. The model includes no volatility, fees, taxes, mortality, changing asset assumptions, or later worker cohorts. A payout that exceeds the available reserve is reported as unfunded rather than producing a negative bucket balance.

Single-year mode displays the contribution, return, payout, and workers for the selected year. Cumulative mode aggregates those flows from 2024 through the selected year; it does not add new cohorts or contributions. In both modes, bucket cards and the final waterfall bar are closing balances as of the selected year-end. Before the final retirement year, those assets still support future inflation-adjusted refunds and are not a free surplus. After all modeled obligations have been paid, the remaining reserve is the modeled cohort surplus. The waterfall is not a forecast of trust-fund solvency or net fiscal benefit.

Worker-count benchmarks

The raw CPS estimate of approximately 11.0 million wage-and-salary workers at or above the cap is used as the pilot population. Survey weights can produce a fractional population estimate, so the editable default is rounded to the nearest whole estimated worker: 10,991,142. It is comparable in concept to SSA’s 11.523 million workers at the taxable maximum in 2023. SSA’s separate 22.119 million figure counts workers with self-employment earnings and must not be interpreted as workers above the cap.

Important limitations

CPS values are survey-reported, public-use high incomes are disclosure protected, and CPS earnings do not perfectly match OASDI-covered administrative earnings. Raw estimates are always retained. Calibration is presented only as a labeled scenario.