Assumptions & Methodology

Backed by US Treasuries; uniquely cannot lose value day-to-day

Tracks the Bloomberg US Aggregate Bond Index. Long-run historical average ~3.5%; recent 10-year average ~1.8% due to 2022 bond correction. RetireCiv uses the long-run figure for planning.

Tracks the S&P 500

Tracks the Dow Jones US Completion TSM Index

Tracks the MSCI EAFE Index (developed international markets)

Applies to all taxpayers

No penalty if you separate from federal service in or after the year you turn 55

Applies to special provision employees who separate at or after age 50

Effective Dec 29, 2022. Public safety employees (LEO, FF, ATC, CPO) who retire at any age with 25+ years of qualifying public safety service are exempt from the 10% penalty

Applied to distributions before penalty-free age unless exception applies

SECURE 2.0 Act; age 73 for those born 1951–1959, rising to age 75 for those born in 1960 or later

Treasury Reg. §1.401(a)(9)-9(c), T.D. 9930. RetireCiv enforces the RMD floor in drawdown: the retiree always withdraws at least the table amount once they reach the start age, even if it is not needed for expenses.

Only traditional TSP, traditional IRA, and 401(k) balances are counted. Designated Roth accounts in employer plans carry no lifetime RMD (SECURE 2.0 §325, effective 2024), Roth IRAs never did, and taxable brokerage accounts sit outside the RMD rules entirely. Your Roth share of the TSP comes from the Taxes and Deductions step; with that turned off the TSP is treated as fully traditional. A spouse’s TSP is also treated as fully traditional.

An RMD is a forced distribution, not forced spending. Whatever you do not need that year is returned to your portfolio as taxable savings, and it is not subject to RMDs again. It is not modeled as a separately allocated taxable account, so its yearly dividends and realized gains are not taxed as they accrue.

Monthly contributions are added after each month’s return is applied, mirroring actual payroll deposits. Prior end-of-period modeling is corrected.

Applies to most FERS employees at retirement

Requires age 62+ AND 20+ years of service at retirement. Unused sick leave counts toward the 20-year threshold (OPM BAL 18-103), though never toward retirement eligibility itself

Benefits continue anyway if you were married at least 30 years and the remarriage happened after January 1, 1995. We cannot know whether a survivor will remarry.

Age 50 if your spouse has a disability. Same reason.

Our age-60 floor is too strict for this case. We do not collect children.

Both need beneficiary or disability details we do not hold.

Applies if separated from service at 50+

Hourly rate multiplied by unused annual leave hours at separation

Civilian FERS service only, rounded to the nearest whole year (5 U.S.C. §8421(b)(1); OPM RI 90-8). Bought-back military time and CSRS-component years count toward eligibility and the annuity, but are excluded from this fraction. RetireCiv subtracts them when you provide them in the wizard.

VERA retirees who leave before MRA must defer SRS to MRA per 5 CFR §842.505; enforced in calculations.

5 U.S.C. §8421a. RetireCiv reduces projected SRS by $1 for every $2 of user-entered earned income above the annual SSA exempt amount.

5 U.S.C. §8421a(c). Special-provision retirees are not subject to the earnings test until they reach their Minimum Retirement Age; RetireCiv pays their SRS in full below MRA regardless of earnings.

SSA Program Explainer: WEP, ELY 2024. Repeal retroactive to Jan 2024; no new reductions apply.

Social Security Fairness Act, P.L. 118-273, signed Jan 5, 2025

SSA sent payments starting Feb 25, 2025. Check SSA.gov for your payment status.

The projection assumes your investments earn the same return every single year. Real markets do not. A plan that looks comfortable here can still fail in a bad decade, which is exactly what the Monte Carlo page is for: it runs thousands of varied market paths and reports how often the plan survives.

The same rate the dashboard uses, adjustable in Settings. Because there is no volatility, it is applied without variation rather than as an average.

A separate control, because a portfolio you are drawing on is usually held more conservatively than one still growing. It starts at 4%, the midpoint of the Monte Carlo page’s post-retirement range, rather than the higher rate assumed while you are working. Only the drawdown years change; your balance on the retirement date is unaffected.

The app’s single long-run CPI figure, adjustable in Settings, applied to expenses and to COLA-bearing income.

Spending is the plan variable on the what-if page. You set it as the share of your current household income you expect to live on in retirement, which is how most retirement guidance is written; published ranges land between 55% and 80%. The slider opens at 80%, the top of the 55% to 80% band cited under the control, rather than at your entered-expense share, which understates a real retirement. Moving it is your decision. Your figure is in today’s dollars and is inflated to each future year. It drives how much comes out of your savings, when they run out, and what is left at the end.

When you have entered expenses, the chart draws a dashed line at that figure, inflated the same way. It does not move when you change the spending slider, so it stays a fixed reference for how far a plan sits from your current life. Loan payments are excluded from it, because they end at their payoff dates. Your mortgage is handled the same way, in two parts: the principal and interest stop at the projected payoff date, and until then they are held flat rather than inflated, because a fixed-rate payment does not rise with prices. Property tax and insurance carry on and do inflate. With no expenses entered the line is not drawn at all rather than sitting at zero. The spending level you chose is not drawn as its own line, because the stacked bars already add up to it.

When your guaranteed income (annuity, Social Security, pensions) is larger than the spending level you chose, the difference is added to your savings and grows there for the rest of the projection. It is treated as taxable savings, so it is never subject to required minimum distributions later. This is why spending less leaves more behind even in years when you withdraw nothing at all. Only the money you could actually keep is saved: we first subtract an estimate of the federal income tax owed on that income. Without that step the balance would grow by dollars that belong to the IRS. The estimate assumes no state income tax, so it is deliberately on the low side.

The projection runs in future dollars, so the closing balance is stated in the money of the year it is reached, not in today’s money. Decades of inflation sit between the two: a figure at age 90 buys considerably less than the same number does now.

The projection withdraws what your plan actually needs: the gap between your spending and your guaranteed income, with the required minimum distribution as a floor. A withdrawal-rate setting would be a second, competing driver for the same number, so the what-if page does not have one. Your saved 4% assumption still drives the dashboard.

Because withdrawals are sized to cover the gap, total first-year income moves in step with whatever spending level you choose: raising the percentage raises the total, but only by drawing more from your own savings. Guaranteed income (FERS annuity, SRS, Social Security, VA compensation, and pensions) does not move with the slider, so it is the figure that tells you what the plan itself produces.

Every income figure on the what-if page is before tax. Your dashboard and the Monte Carlo page both offer an after-tax view when the Taxes and Deductions estimate is turned on. There is one exception, and it runs in your favour as a matter of accuracy: unspent income is taxed before it is added to your savings (see the row above). A balance is not a before-tax figure the way an income figure is, so growing it with untaxed money would simply overstate it.

The long-run diet-COLA average by default, adjustable in Settings. The chart and the summary cards now use the same rate, so the annuity band and the figures above it always agree.

The projection ends at the age you choose. The final year shown is the year before that age, since the year of death pays no full year of income. A spouse who outlives you keeps the projection running on survivor income.

What-if settings live only in the page while you are on it. They never change your saved plan and are never sent to a server.

Balances and incomes reported by the simulation are in the dollars of the year they occur, not today’s money. A median ending balance decades out therefore looks larger than its purchasing power. The what-if projection states its own basis separately.

You set the range on the Monte Carlo page. The simulation takes the midpoint as the mean and the half-width as the standard deviation, so the default range runs at a 7% mean with a 3% standard deviation. Growth-oriented allocation (equity-heavy, e.g., TSP L 2040+). Sampled monthly at mean/12 and std/√12 to preserve annualized parameters while capturing monthly sequence risk.

Set on the Monte Carlo page; midpoint is the mean, half-width the standard deviation, so the default range runs at a 4% mean with a 2% standard deviation. Conservative allocation (bond-heavy, e.g., TSP L Income). Same monthly aggregation.

Set on the Monte Carlo page; midpoint is the mean, half-width the standard deviation, so the default range runs at a 2.5% mean with a 1.0% standard deviation. Anchored on the SSA Trustees Report long-run CPI-W projection. Drawn per simulated year, clamped at ≥ −2% to prevent runaway deflation.

FERS COLAs are capped below CPI by the diet-COLA rule (5 U.S.C. §8462(b)): if CPI ≥ 3% → CPI − 1%; 2%–3% → 2%; < 2% → full CPI. The simulation models the long-run average as a flat 1.8%. Regular FERS receives no COLA until age 62; special-provision (LEO/FF/ATC) and disability retirees begin COLA at retirement.

CPI-W pass-through with SSA’s statutory 0% floor (benefits do not decrease in deflationary years).

Most private, state, and military pensions have no COLA. Users with a COLA-bearing pension can supply a rate.

IRS Uniform Lifetime Table (Treasury Reg. §1.401(a)(9)-9(c)), applied to tax-deferred balances only. Drawdown withdraws at least the RMD amount from your start age onward, even if expenses would not require it; anything left unspent is returned to the portfolio as taxable savings.

5 U.S.C. §8421a. SRS reduced $1 per $2 of earned income above the annual SSA exempt amount. Special-provision (LEO/FF/ATC) retirees are exempt until they reach their Minimum Retirement Age.

Your lifespan cap comes from your own slider. With the survivor analysis on, the simulated horizon extends to the later of the two deaths so surviving-spouse years are still modeled.

Guards against log(0) → -Infinity edge case that would contaminate a draw with NaN.

When a seed is supplied via simulationParams.seed, every iteration shares a deterministic sequence. Omit the seed for production runs backed by Math.random.

PRE/POST_RETIREMENT_PARAMS are nominal. Inflation is drawn independently per year and applied to expenses, COLAs, and actual-expense inflation using the same within-year draw.

Spending keeps pace with inflation for life. This is what Bengen assumed in the research behind the 4% rule, and what most financial planning software still assumes.

A flat, easy-to-explain rate. Close to what Hurd and Rohwedder measured across every wealth quartile in the Health and Retirement Study, and the same simplification RightCapital ships in its own planning software.

Blanchett (2026), Equation 2. Tracks the median retiree, whose real spending keeps falling at older ages. The formula is slightly positive below age 63, so if you retire before then your spending rises a little first. It turns down around 63 and falls fastest in the 80s.

Blanchett (2026), Equation 1, unchanged from his 2014 paper. Tracks the average across all retirees. The formula is positive below age 65, so an early retiree spends more in real terms for the first several years (see the row below). After that it declines, and while it flattens in the late 80s, the upturn never restores your original standard of living: it only slows the decline.

Peak real spending compared with your first year, once we apply the pattern to the flexible 75%: retire at 62 and the smile peaks about 1% above day one, which is barely visible. Retire at 57, the earliest most FERS employees can, and it peaks about 6% above. Retire at 50 under the law enforcement or firefighter provisions and it peaks about 13% above, around age 66. The smirk is much smaller: under 1% at 62, about 3% at 50. This is what the published formulas say, not an adjustment we added.

The smile is the average retiree; the smirk is the median one. The gap between them is health shocks, which move an average without moving a median. Blanchett’s own conclusion is that moving off constant spending matters far more than which declining shape you pick.

Both formulas are second-order polynomials fitted to that range, and neither is valid outside it. We hold the rate flat below 60 and above 90. Extrapolating the smile to age 105 would produce a real spending increase of about 4.4% a year, which is the opposite of the finding it comes from.

A decline does not fall evenly on a budget. Following Blanchett’s own essential-versus-flexible split, we apply the pattern to 75% of your spending and hold the remaining 25% flat in real terms. Your mortgage principal and interest are excluded entirely, on top of that: a fixed-rate payment does not shrink because you got older, and it stops at payoff.

FEHB, FEDVIP, and Medicare premiums are handled separately in the optional taxes and deductions estimate, where they grow at your healthcare inflation assumption. They are the one part of a retiree budget with strong evidence of real growth, so no spending pattern is ever applied to them.

The survivor spending level and the spending pattern are separate assumptions and they multiply. We apply the pattern first, then the survivor level. They do not double-count: the research only includes households whose marital status stayed the same across every survey, so a spouse dying is not part of the measured decline. It is a genuinely separate adjustment.

The formulas come from a filtered sample. A household had to answer every survey, be fully retired, keep the same marital status throughout, spend between $25,000 and $500,000, and avoid very large swings between surveys. In the 2014 study that left about 11% of available households. The pattern describes settled retirements, so treat it as a reasonable guide rather than a forecast of your own budget.

That real spending falls is well supported and found by several independent researchers. The precise curve is less certain. A 2026 replication by Derek Tharp found that when you follow the same households over time, rather than comparing different households at different ages, the smile’s curve is not statistically distinguishable from a steady decline. That is a reason to prefer the simple gentle decline if you want the most defensible option, and a reason we still default to no decline at all.

Your readiness score and replacement rate ignore whichever pattern you pick, the same way they always ignore taxes. A headline score should not improve because you chose a friendlier assumption.

Varies by entry year: 0.8% (regular) / 3.1% (FERS-RAE) / 4.4% (FERS-FRAE)