RetirementForge

Calculation methodology

Where every number comes from, and where it stops.

Thirty calculators is an easy claim to make. Thirty calculators you can defend to a sophisticated client is a different one. This page lists the authority behind every government-published figure in the platform, when each is next due for review, and — engine by engine — exactly what we do not model.

The figures on this page are not typed here. They are imported.

Every value in the table below is read from the same constant the calculator itself uses. Nothing is transcribed, which means there is no version of this site where the published figure and the calculated figure disagree — the page and the engine are reading one number.

That is a deliberate design choice rather than a boast. We had the provenance of these figures written down in three separate places, and one of them drifted: our internal documentation was still describing the 2025 Medicare schedule and a $185.00 Part B premium months after the engine had moved to the current CMS figures. Nothing was wrong in the calculator. The description of it was wrong, which is its own kind of failure, and copying that description onto a public page would have made it worse.

The figures, and who publishes them

23 figures across 6 areas. A dash under Published means the authority issues the figure without a citable document date.

Federal income tax

FigureAuthorityPublishedNext review
Ordinary income brackets (2026)Seven brackets, 10%–37%. The 22% bracket tops at $105,700 for a single filer.IRSRev. Proc. 2025-32 § 4.01, Tables 1–42025-10-09October
Standard deduction (2026)$16,100 single · $32,200 married filing jointly · $24,150 head of householdIRSRev. Proc. 2025-32 § 4.012025-10-09October
Additional standard deduction, age 65+ (IRC § 63(f))$2,050 if unmarried and not a surviving spouse · $1,650 per qualifying individual if marriedIRSRev. Proc. 2025-32 § 4.14(3)2025-10-09October
Temporary senior bonus deduction (OBBBA)Available for tax years 2025–2028 only, and stacks on top of the § 63(f) amount above.Scheduled to lapse after its final year. The engine applies it only inside that window rather than assuming an extension.CongressOne Big Beautiful Bill Act—Statutory

Medicare & IRMAA

FigureAuthorityPublishedNext review
Standard Part B premium$202.90/monthCMSFact sheet, "2026 Medicare Parts A & B Premiums and Deductibles"2025-11-14November
IRMAA surcharge tiers5 surcharge tiers above the standard premium. At the top tier a single filer pays $578/month in combined Part B and Part D surcharges — $6,936 a year.CMSFact sheet, "2026 Medicare Parts A & B Premiums and Deductibles"2025-11-14November

Contributions & HSA

FigureAuthorityPublishedNext review
Elective deferral limit — 401(k) / 403(b) / TSP (IRC § 402(g))$24,500IRSNotice 2025-67—January
Catch-up contributions$8,000 at age 50+ · $11,250 "super catch-up" at ages 60–63, which replaces the age-50 amount rather than adding to it.IRSNotice 2025-67; SECURE 2.0 § 109—January
Wage threshold forcing catch-ups to Roth (SECURE 2.0 § 603)$150,000 in prior-year FICA wagesIRSIRC § 414(v)(7)(A)—January
Overall annual additions limit (IRC § 415(c))$72,000, employee plus employer. Catch-up contributions sit on top of this.IRSNotice 2025-67—January
HSA contribution limits (IRC § 223(b))$4,400 self-only · $8,750 family · $1,000 catch-up at 55+IRSRev. Proc. 2025-19—January
Qualifying HDHP thresholds (IRC § 223(c)(2)(A))Minimum deductible $1,700 self-only / $3,400 family · maximum out-of-pocket $8,500 / $17,000IRSRev. Proc. 2025-19—January
Eligible long-term care premiums (IRC § 213(d)(10))Age-banded, from $500 at 40 and under to $6,200 above 70.IRSRev. Proc. 2025-32 § .272025-10-09October

ACA premium tax credits

FigureAuthorityPublishedNext review
Federal poverty level$15,650 for a one-person household · $5,500 for each additional personHHSHHS poverty guidelines, 90 Fed. Reg.2025-01-17January
Applicable percentage table (IRC § 36B)6 income bands, from 2.10% of household income at the bottom to 9.96% at the top.IRSRev. Proc. 2025-25 § 3.01—Summer
Premium tax credit cliffHousehold income above 400% of the federal poverty level receives no credit at all.The enhanced credits from ARPA and the Inflation Reduction Act expired and were not renewed, so the cliff is back. This one is reviewed alongside the applicable percentage table because it turns on legislation rather than an indexed figure.CongressIRC § 36B, as the enhanced credits lapsed after 2025-12-31—Summer

Social Security & payroll

FigureAuthorityPublishedNext review
PIA bend points$1,286 and $7,749 of monthly indexed earnings — the points where the benefit formula steps down from 90% to 32% to 15%.SSABend points for workers first eligible in 2026—October
Social Security wage base$184,500SSAContribution and benefit base, announced with the annual COLA—October
FICA rates (employee share)6.20% Social Security up to the wage base · 1.45% Medicare with no wage capThe rates themselves are set in statute and do not move with inflation. Only the wage base they apply to is indexed, and that is tracked separately above.IRSTopic No. 751—Statutory
Additional Medicare Tax0.90% on wages above $200,000 single / $250,000 married filing jointlyStatutory and deliberately not indexed — these thresholds have been unchanged since 2013. Listing them as due for annual review would imply a movement that does not happen.CongressIRC § 3101(b)(2)—Statutory

Required distributions

FigureAuthorityPublishedNext review
Age required minimum distributions begin72 for those born 1950 or earlier · 73 for 1951–1959 · 75 for 1960 and laterSet by statute rather than indexed. Birth year 1959 was genuinely ambiguous as drafted — the statute read as both 73 and 75 — and we follow the IRS position from its July 2024 proposed regulations. The final regulations reserved that paragraph, so this is the figure on this page most likely to change.CongressSECURE 2.0 § 107—Statutory
Uniform Lifetime Table (lifetime RMD divisors)Ages 72–120. The divisor at 75 is 24.6; at 90 it is 12.2.A regulation table rather than an annual figure. It changes only when Treasury amends the regulation, which last happened for distribution years beginning in 2022.IRSTreas. Reg. § 1.401(a)(9)-9(c), as amended by TD 99302020-11-12Statutory
Single Life Table (inherited account divisors)Used for beneficiary distributions. The factor at 55 is 31.6.A regulation table. The July 2024 final regulations confirmed the annual-distribution requirement for beneficiaries subject to the 10-year rule, which the engine applies.IRSPublication 590-B, Appendix B, Table I—Statutory

How it stays current

The agencies do not publish on one schedule, so neither do we.

January

The IRS publishes inflation-adjusted contribution and HSA limits, and HHS issues the year's poverty guidelines.

7 figures reviewed: Elective deferral limit — 401(k) / 403(b) / TSP (IRC § 402(g)) · Catch-up contributions · Wage threshold forcing catch-ups to Roth (SECURE 2.0 § 603) · Overall annual additions limit (IRC § 415(c)) · HSA contribution limits (IRC § 223(b)) · Qualifying HDHP thresholds (IRC § 223(c)(2)(A)) · Federal poverty level

Summer

The IRS publishes the following year's § 36B applicable percentage table.

2 figures reviewed: Applicable percentage table (IRC § 36B) · Premium tax credit cliff

October

The IRS releases the annual Revenue Procedure of inflation-adjusted tax items, and SSA announces the COLA, the wage base and the new bend points.

6 figures reviewed: Ordinary income brackets (2026) · Standard deduction (2026) · Additional standard deduction, age 65+ (IRC § 63(f)) · Eligible long-term care premiums (IRC § 213(d)(10)) · PIA bend points · Social Security wage base

November

CMS publishes the coming year's Medicare premiums and IRMAA schedule.

2 figures reviewed: Standard Part B premium · IRMAA surcharge tiers

And 6 figures with no review month at all

These are statutory rather than inflation-indexed — they move only when Congress moves them. The Additional Medicare Tax thresholds, for instance, have been unchanged since 2013. Listing them as due for annual review would imply a movement that does not happen, and a review calendar full of items that never change is a calendar nobody reads.

Why we audit the rules, not just the numbers

A figure going stale is the easy failure. It is visible, a test can pin it, and a calendar catches it. The failure that actually costs an advisor credibility is subtler: the arithmetic stays perfect while the regime underneath it disappears.

Our ACA engine did exactly that. It modeled the enhanced premium subsidies from the American Rescue Plan and the Inflation Reduction Act — correctly, precisely, and for roughly ten months after those subsidies had expired. Every internal check passed. The tests passed too, because tests pin the behaviour they were written against; they faithfully protected the wrong regime. For those ten months the tool told advisors their clients kept help those clients no longer received.

We found it in a structured audit, fixed it, restored the 400% cliff, and pinned the corrected behaviour to the Revenue Procedure that publishes it. We are telling you about it because the alternative — describing our process in the abstract and hoping you assume it has never caught anything — would be the less honest option, and because every vendor you evaluate has a story like this one. Most will not offer it.

So each engine review now asks two questions rather than one. Are the numbers right? And are the rules still the rules? The second question is the one that found Nebraska’s impossible tax bracket, a public FAQ that contradicted its own calculator, and a pension recommendation that was inverted on its own default inputs — none of which any test would have caught.

What each engine models — and what it does not

The second list is the useful one. It tells you where the screen stops and your own judgment starts.

Federal income tax

The bracket engine every other tool leans on when it needs to know what a dollar of income costs.

What it models

  • All seven ordinary brackets for single, married filing jointly, married filing separately, and head of household, from the current-year Revenue Procedure.
  • The standard deduction, the § 63(f) additional deduction at 65, and the temporary OBBBA senior bonus deduction — including the fact that the last one stacks on the first and lapses after its final year.
  • Forward projection with inflation indexing, so a multi-year plan does not apply this year's brackets to a client fifteen years out.

What it does not

  • The Net Investment Income Tax of 3.8% is not applied. It rarely changes a decision on its own, but a conversion that raises modified AGI can expose a client's other investment income to it.
  • The Alternative Minimum Tax is not calculated.
  • The § 63(f) allowance for blindness is not modeled — the engine has no blindness input, only age.
  • State tax is handled by a separate module with its own simplifications, not by this one.

src/lib/federal-tax/

Medicare IRMAA

Part B and Part D surcharges by income tier, and the effective marginal rate a Roth conversion pays for crossing one.

What it models

  • Every surcharge tier for single and joint filers on the current CMS schedule, for both Part B and Part D.
  • The effective IRMAA rate on a marginal conversion dollar — the additional annual cost divided by the conversion amount — which is the number that actually decides whether a conversion is worth it near a threshold.
  • Tier ceilings, so the multi-year Roth projection can hold conversions under a chosen cap.

What it does not

  • The standalone calculator does not model the two-year lookback. It answers "what surcharge does this MAGI produce", not "a conversion this year raises the premium two years from now". The multi-year Roth projection does apply the timing; the quick calculator does not.
  • Life-changing event appeals on Form SSA-44 — retirement, work stoppage, divorce, death of a spouse — are not modeled. A client who qualifies may pay far less than the tool shows.
  • Household income is not split between spouses for optimization; joint filers are treated as one MAGI.

src/lib/irmaa/

Roth conversion projection

A year-by-year comparison of a conversion strategy against doing nothing, across the full retirement window.

What it models

  • Federal brackets with annual inflation indexing, state tax for the client's state, and the § 86 provisional-income formula for taxation of Social Security.
  • Required minimum distributions under SECURE 2.0, and IRMAA surcharges on the two-year lag.
  • After-tax IRA basis recovered pro-rata across distributions, following Form 8606.
  • The 10% early-distribution penalty when conversion tax is paid from the IRA before 59½.
  • Legacy outcomes for heirs, including an explicit ten-year distribution schedule under the July 2024 final regulations when an heir profile is supplied.

What it does not

  • Neither five-year clock is applied. Roth withdrawals are modeled as entirely tax-free, so the rule for qualified distributions of earnings and the separate recapture rule on converted amounts withdrawn before 59½ are both absent. A client whose first Roth account is opened by one of these conversions needs both clocks checked by hand.
  • The Net Investment Income Tax is not modeled.
  • Statutory exceptions to the early-distribution penalty — 72(t) substantially equal payments, disability — are not applied.
  • Employer plans are assumed to hold no after-tax basis.
  • Charitable strategies are absent. Qualified charitable distributions in particular can materially reduce RMD-driven taxable income, and the engine will not show that.
  • Returns are applied uniformly each year. Sequence-of-returns risk lives in the Monte Carlo tool, not here.

src/lib/roth/

Social Security claiming

Break-even ages across claim strategies, and what happens to a household when the first spouse dies.

What it models

  • Benefits at 62, full retirement age, and 70, from the current bend points, with COLA-adjusted cumulative totals and the break-even age between each pair.
  • Household survivor mechanics: which check stops, which continues, and the tax-filing change that follows.
  • Pre-repeal WEP and GPO offsets in the advanced panel, for advisors recalculating a client against the prior environment.

What it does not

  • The SSA earnings test is not applied. A client claiming early while still working will see real withholding the tool does not show.
  • Claim-age optimization is not continuous. The comparison covers discrete scenarios rather than sweeping every month between 62 and 70.
  • Full retirement age is a binary toggle of 66 or 67, so the partial-month ages for birth years 1955 to 1959 are approximated.
  • First death is a fixed age the advisor enters, not a mortality distribution. For probabilistic outcomes, pair it with the Monte Carlo tool.
  • The survivor is assumed to claim the survivor benefit immediately at first death. Taking it before the survivor's own survivor FRA carries reductions not modeled here.
  • The WEP reduction uses the 50%-of-pension guarantee cap rather than the modified bend point formula. That is the binding constraint in most real cases, but it understates the reduction for very-low-AIME workers.

src/lib/social-security/

Required minimum distributions

Lifetime RMDs for account owners and the ten-year rule for beneficiaries.

What it models

  • The Uniform Lifetime Table from the Treasury regulation, ages 72 through 120, held in one place and shared by every engine that needs it.
  • The SECURE 2.0 applicable age by birth year.
  • The Single Life Table for inherited accounts, and the annual-distribution requirement the July 2024 final regulations confirmed for beneficiaries subject to the ten-year rule.

What it does not

  • The Joint and Last Survivor Table is not implemented. An owner whose sole beneficiary is a spouse more than ten years younger uses a different, more favorable table, and this tool will overstate their RMD.
  • Birth year 1959 is genuinely ambiguous in the statute — it reads as both 73 and 75. We follow the IRS position from its July 2024 proposed regulations, but the final regulations reserved that paragraph. This is the single figure in the platform most likely to change.
  • Actual RMDs can differ where beneficiary designations are unusual.

src/lib/rmd/, src/lib/inherited-ira/

ACA premium tax credits

What a pre-Medicare client pays for marketplace coverage, and what income change costs them the credit.

What it models

  • The § 36B applicable percentage table with linear interpolation inside each income band.
  • The 400% federal poverty level cliff, which is back for 2026 now that the enhanced credits have lapsed.
  • Year-by-year projection, so an advisor can see the year a conversion pushes a client over.

What it does not

  • Benchmark plan premiums are an advisor input, not a live quote from the marketplace. Actual second-lowest-Silver pricing varies by county and changes annually.
  • Cost-sharing reductions below 250% of the poverty level are not modeled — only the premium tax credit.
  • Income below 100% of the poverty level is treated as producing no marketplace subsidy, on the assumption of Medicaid eligibility. Actual state Medicaid rules are not applied, and the coverage gap in non-expansion states — where a household is too poor for a credit and still ineligible for Medicaid — is not modeled.

src/lib/aca/

Contribution limits and HSA

What a client can put away this year across every plan type, and what an HSA is worth against the alternatives.

What it models

  • The § 402(g) elective deferral limit shared across 401(k), 403(b) and TSP, the separate 457(b) limit that allows a genuine double-dip, and the § 415(c) overall additions cap.
  • The SECURE 2.0 super catch-up at ages 60 to 63, including the fact that it replaces the age-50 catch-up rather than adding to it, and the § 603 wage threshold that forces catch-ups to Roth.
  • HSA limits and HDHP qualification thresholds, the six-month retroactive Part A lookback that catches savers enrolling in Medicare, and the § 213(d)(10) age-banded caps on eligible long-term care premiums.

What it does not

  • The 403(b) fifteen-year service catch-up is modeled on its statutory caps, but eligibility depends on an employer service history the tool does not hold.
  • Plan-specific limits below the statutory maximum — a plan document capping deferrals, or a plan that does not permit catch-ups — are not known to the tool.
  • Nondiscrimination testing outcomes that can force a refund of deferrals after year end are not modeled.

src/lib/contribution/, src/lib/hsa/

Monte Carlo simulation

Probability of a portfolio lasting, across many randomized return paths.

What it models

  • Randomized annual returns across the projection window, with guaranteed income streams handled separately from the volatile portfolio.
  • A seed derived from the inputs themselves, so the same scenario always produces the same answer. Running the tool twice in front of a client cannot produce two different probabilities.

What it does not

  • Stocks and bonds are drawn with zero correlation between them. Real correlation is not zero and has been markedly positive during inflationary periods, which means the simulation understates the chance of both falling together — the scenario that most damages an early retiree.
  • Seeding fixes reproducibility, not accuracy. A stable number is not a more correct number.
  • Return distributions are drawn from advisor-set assumptions rather than a historical bootstrap, so the output is only ever as good as the assumptions on the screen.

src/lib/monte-carlo/

Figures are pinned to their source, not to their output

A test that asserts a calculator returns what it currently returns proves nothing except that nobody has changed it. Our figure tests cite the instrument instead — the Revenue Procedure, its section, its publication date — so a future edit that moves a threshold fails against the authority rather than quietly redefining what correct means. Regulation tables like the Uniform Lifetime Table are diffed row by row against the published table rather than spot-checked at the ages a screen happens to display.

Frequently asked questions

How often does RetirementForge update its tax and Medicare figures?
On the schedule the agencies publish, not on a fixed annual sweep. The IRS releases inflation-adjusted tax items each October in a Revenue Procedure, SSA announces the COLA, wage base and bend points the same month, CMS publishes Medicare premiums and the IRMAA schedule in November, HHS issues poverty guidelines in January alongside the IRS contribution and HSA limits, and the § 36B applicable percentage table arrives in summer. Each figure in the platform carries the month it is due for review, and figures that are statutory rather than inflation-indexed carry no review month at all, because they move only when Congress moves them.
How do I know the numbers on the methodology page match the calculators?
Because they are the same numbers. Every value in the figures table is imported from the identical constant the calculator uses at runtime, rather than typed into the page by hand. There is no version of this site where the published figure and the calculated figure disagree, because there is only one figure. Updating the engine updates the page.
Are the calculations tested?
Yes, and the tests pin figures to the document that published them rather than to whatever the code happened to produce. A test file for the 2026 federal brackets, for example, cites the Revenue Procedure and its publication date, so a future edit that changes a threshold fails against the source rather than silently redefining what is correct.
What does RetirementForge deliberately not model?
A number of things, listed by engine on this page rather than buried in a disclaimer. Among them: the Net Investment Income Tax and the Alternative Minimum Tax, both five-year Roth clocks, the SSA earnings test, IRMAA life-changing-event appeals on Form SSA-44, the Joint and Last Survivor Table for RMDs, and correlation between stocks and bonds in the Monte Carlo simulation. Each one is a place where an advisor needs to know the screen stops and their own judgment starts.
Does RetirementForge give investment advice?
No. The platform performs calculations on inputs an advisor supplies and displays the results. It does not recommend securities or products, and its output is educational and for planning discussion rather than a recommendation. The advisor using it is the one exercising judgment, and remains responsible for it.
What happens when tax law changes mid-year?
A figure moving is the easy case and the review calendar catches it. The harder case is a rule changing — a provision expiring, a threshold being repealed, a structure being rewritten — because nothing errors when that happens. The engine keeps producing confident numbers under a regime that no longer exists. That is why each engine review asks two separate questions: whether the numbers are right, and whether the rules are still the rules.

Found something wrong? We would rather know.

If a figure on this page disagrees with the authority that publishes it, or an engine models a rule that has since changed, tell us and we will correct it and say so. Nothing on this page is worth more than its accuracy.

support@retirementforge.com

See also our Security & Data Protection page and the public calculators.