Where the formulas come from
Each plan's constants — percentages, brackets, poverty-guideline tables, terms — were transcribed from official publications, and every constant in the engine carries its source and retrieval date. The primary sources:
- RAP: Federal Student Aid's RAP information center, cross-confirmed by the Congressional Research Service summary of the statute.
- IBR, PAYE, ICR: the official IDR application's plan details and 34 CFR 685.209; ICR's 2026 income percentage factors from 91 FR 34815 (the June 2026 Federal Register update).
- Poverty guidelines: the 2025 and 2026 HHS tables from ASPE and the Federal Register, all three state groups, all family sizes.
- SAVE wind-down: the official SAVE court actions page.
The rules implementing RAP and Tiered Standard were published 2026-05-01 and are under active litigation. That's why a visible "Rules as of" stamp sits on every page: if a court or a new rule changes a value, the stamp tells you which version produced your numbers.
The formula for each plan
- RAP: annual payment = a bracket percentage (1–10%) of your entire AGI, ÷ 12, − $50 per tax dependent, floored at $10/month. Unpaid interest in an on-time month is waived; principal drops by at least $50/month via a matching credit; forgiveness after 360 payments. Brackets use "more than X, not more than Y" boundaries — exactly $100,000 is the 9% bracket. The full table is here.
- IBR: 10% (first loan on/after July 1, 2014) or 15% (earlier) of discretionary income — AGI minus 150% of the poverty guideline for your family size — ÷ 12, capped at the 10-year Standard amount, forgiveness after 20 or 25 years respectively.
- PAYE: 10% of the same discretionary income, capped at the 10-year Standard amount, forgiveness after 20 years; payments under $5 bill at $0 and $5–$10 bill at $10. Eligibility requires no federal balance as of Oct 1, 2007 and a Direct Loan disbursement on or after Oct 1, 2011. Retired no later than July 1, 2028.
- ICR: the lesser of (a) a 12-year amortization of your balance multiplied by an income percentage factor interpolated from the published table, or (b) 20% of AGI minus 100% of the poverty guideline. No Standard cap; forgiveness after 25 years counting only payments through July 1, 2028. Our implementation reproduces the Federal Register notice's own worked examples to the cent.
- Standard 10-year: level payments that amortize the balance. For borrowers already in repayment, official tools quote the current balance over the months remaining on the original schedule at the principal-weighted rate — so when you provide your servicer's scheduled payoff date, RepayCompass does the same; without it, a fresh 120-month amortization.
- Tiered Standard: level payments over 10/15/20/25 years by balance tier (<$25k / $25–49,999 / $50–99,999 / ≥$100k), $50 monthly minimum, only for borrowers with a loan or consolidation on or after July 1, 2026.
Married borrowers and household math
When you are married, file jointly, and both you and your spouse carry federal loans, the income-driven plans do not simply add two separate payments together. Under 34 CFR 685.209(g), each spouse's RAP, IBR, or PAYE payment is prorated to that spouse's share of the couple's debt. RepayCompass computes the payment under the plan's normal formula first — for IBR and PAYE, after the 10-year Standard cap — then multiplies it by your share: your loans' outstanding principal and interest ÷ the couple's combined outstanding principal and interest. The rule spells this out for RAP at (g)(3)(i) and for IBR and PAYE at (g)(1)(i); a spouse's debt is counted only because the spouse's income is, and the two travel together under (e)(2)(i).
A worked example. Suppose the two of you owe $100,000 between you, and $60,000 of that is yours. Your share is 60% and your spouse's is 40%. If the household RAP figure comes to $500 a month, you are billed $300 and your spouse $200 — 60% and 40% of one number, not two unrelated calculations.
Two cases sit outside proration. ICR is never prorated — the spousal-debt adjustment in (g)(1) leaves it out — and filing separately never prorates either, because a separate return keeps your spouse's income, and so their debt, out of your payment entirely.
The share moves as you both pay down. The rule recomputes it at each annual recertification from whatever the two of you owe then, and RepayCompass projects it the same way. Your first payment always uses today's balances — that is the figure official tools quote — and from year one on the split follows the balances. Whoever still owes after the other finishes carries the whole household payment.
Two things worth knowing about that. Because each payment is a share of one household amount, the spouse with more debt pays proportionally more and the two balances tend to run down together, so a couple often finishes closer to the same time than the separate figures suggest. And because your spouse's balance depends on which plan they choose, each of your plan cards assumes your spouse is repaying under that same plan.
Plan pairings. The household summary also compares plan pairings — both of you on RAP against you on RAP and your spouse on IBR — ranked on combined lifetime cost: payments plus estimated tax on any forgiveness. A mixed row does not inherit the cards' same-plan assumption: the couple's balance split is re-solved with each spouse on that row's own plan. Two limits are built in, and the table discloses both. Solving a pairing takes real computation, so only a shortlist of the mixed combinations is priced — picked by combined first payment, which is not the metric being compared, so a pairing that costs more now but less over time can be missed, and the table says how many combinations went unpriced. And when both forgiven balances land in the same tax year, the pairing is taxed as one stacked amount on the joint return — more than the two cards' separate estimates added together, because the cards price each discharge alone. If you are both pursuing PSLF, pairings rank on combined cost to your 120th qualifying payments with no tax term, since PSLF forgiveness is not taxed; if only one of you is, your two costs run to different end points and no combined total is shown. Filing separately there is no shared payment to split, so every pairing is exact.
Both filing scenarios, side by side. Because filing status drives the math, the tool computes both. The joint scenario uses your joint AGI, a family size that includes your spouse, and every dependent on the joint return. The separate scenario uses each spouse's own income estimate, drops the spouse from family size (per (b)(9)), and splits the dependents by the claim picker on the form — whichever return each dependent is claimed on. Seeing both lets you weigh a lower separate-filing payment against what filing separately may do to your taxes; that tax trade-off is one to talk through with a tax professional.
The combined cap can exceed your own Standard quote. IBR and PAYE cap the payment at a 10-year Standard amount. For a couple where both spouses have loans, that cap is built from the couple's combined Standard payment — your own Standard quote plus your spouse's — and only then is your share applied. Because Standard payments are not proportional to balance when two people carry different rates or terms, your share of the combined cap can land above what you would pay on your own loans alone under Standard. That reads as surprising, but it is exactly what the regulation's text produces — not a rounding error.
Joint AGI is one number, not a sum. The form asks for the single joint AGI from your joint return — never a your-income field plus a spouse-income field added together, since deductions mean a real joint AGI is almost never the two paychecks summed. The separate own-income fields are estimates used only to build the filing-separately scenario.
How projections work
Every plan is projected month by month over its full term: interest accrues monthly on principal, income-driven payments recompute annually using your income-growth assumption (default 3.15%/year, adjustable 0–8%), and the loop tracks balance, totals, and any forgiven amount. All money math runs in integer cents — rounding happens only at display time.
Where the 3.15% comes from. The official StudentAid.gov calculator grows your income too, but its assumptions page does not say by how much — it documents inflation only for the poverty guidelines and the ICR income-percentage factors. So we solved for it: 3.15% is the rate that reproduces that calculator's own published figures, checked against three very different borrowers. For one it lands within 0.5% on lifetime total and 0.1% on the forgiven amount; for the other two it reproduces the payoff month exactly or within a couple of months. Set it to 0% if you don't expect raises — that is a supported answer, not a broken one. This is an assumption, not a rule, and we show it so you can change it.
If you know your income changes on a date — finishing residency, retiring, a spouse stopping work — you can enter that instead of trying to express it as a growth rate. We hold your income flat-growing until the year you name, switch to the figure you give, and carry on growing from there. Married borrowers enter it per person, because the filing-jointly and filing-separately comparisons need different figures. No official calculator does this, so unlike most of what this tool computes there is nothing to check it against; it is arithmetic on the numbers you supply, not a reproduction of an official quote.
One thing it deliberately does not model: the recertification lag. Your payment follows a real income change only when you recertify, which normally runs on your most recent tax return — so about a year late. We show the change in the year you enter it, because the delay is not fixed: when income falls you can ask your servicer to recalculate on current income and submit documentation such as paystubs instead of authorising the tax-record pull (34 CFR 685.209(l)(2) and (l)(6)). Building in a fixed wait would assert a delay the rules make optional.
Estimated tax on forgiveness is computed from the federal brackets, not a flat rate: the forgiven balance is stacked on top of your projected income for that year and taxed a slice at a time as it climbs, which is how the IRS actually treats it. Because plans forgive different amounts in different years, the effective rate differs from card to card. Brackets and the standard deduction are the verified 2026 figures, inflated forward at the same rate the projection uses for the poverty line — freezing them would invent decades of bracket creep. We assume the standard deduction, since AGI is all we ask for; state tax is not modelled; and a 20–30 year projection of tax law is an estimate, which is why you can override it with a single flat rate. Filing separately has its own bracket table, which is not the single-filer one — it matches at five of seven thresholds and then diverges, so all three filing statuses are computed from their own figures. Shown only when a plan projects forgiveness in 2026 or later. PSLF forgiveness is never taxed and is never given a tax estimate.
Payments you've already made
If you have been repaying for years, a forgiveness clock that started over would overstate what is left. So when you enter your official qualifying-payment count, RepayCompass shortens that plan's forgiveness term by the months that carry to it and re-anchors the projection to today's balance — the remaining time and cost shown are what is left, not a full term beginning again.
The carry is directional. Qualifying payments you have already made carry toward IBR and toward RAP, so switching plans does not reset your progress to zero. RAP months, though, do not carry back to IBR, PAYE, or ICR — that direction runs one way only.
SAVE months split in two. Months you actually paid under SAVE count toward IBR, RAP, and Public Service Loan Forgiveness (PSLF) alike. Months your payments were paused under the SAVE forbearance are treated differently: they do not count toward IBR, RAP, or PSLF. The rule credits only certain named forbearance types — a national emergency, or a short administrative processing window — and the SAVE litigation pause is not one of them. Those paused months need not be lost for PSLF — they may be recoverable through the PSLF Buyback program, but only once you already have 120 certified months of qualifying employment, and never for months you spent on RAP or the Tiered Standard plan — so they are worth confirming with your servicer. That split is why the tool asks for your paid months and your paused months as two separate numbers.
Confirm your counts, and expect these rules to move. The carry rules come from the 2026 RISE final rule, as codified at 34 CFR 685.209 — which is under active litigation and could still change. Your servicer tracks the official counts, which can differ from the number you enter here, so confirm them on StudentAid.gov before relying on these figures.
The "if you consolidate" comparison
Ticking Considering consolidating adds a second comparison below your results, modeled as consolidating today. It changes nothing about the plan cards themselves: those always describe the loans you have now. The what-if is a separate run of the same engine over a transformed profile — your rows merged into one Direct Consolidation Loan — so every plan rule behaves exactly as it does above.
The new loan. Consolidation capitalizes unpaid interest, so the new principal is your balances plus the interest you have accrued. The rate is the weighted average of your rates, rounded up to the nearest higher one-eighth of one percent, per 34 CFR 685.202(a)(10)(i)(F). The regulation does not say what the average is weighted by; we weight by each loan's outstanding balance, which is an assumption rather than a published rule. And because a consolidation made on or after July 1, 2026 can only be repaid under RAP or the Tiered Standard plan, the comparison shows most plans disappearing — that is the lockout working, not a missing calculation.
Payments you've already made: we show both answers, because the rule does not exist. Whether the qualifying payments behind your IDR forgiveness count survive into a new consolidation loan is not settled anywhere we can find. The credit rule that carries those payments across a consolidation applies, by its own words, to the PAYE, ICR, and IBR plans, and a new consolidation loan cannot be repaid under any of them; RAP's forgiveness paragraphs never mention consolidation at all, the Department declined to adopt a rule on the question when asked during rulemaking, and StudentAid.gov's consolidation page says nothing either way. So rather than pick, the panel shows both bounds — what happens if your payments carry, and what happens if the clock restarts — and labels them as exactly that. Where your balance pays off before either clock could matter, the count is moot and the panel says so instead.
PSLF is different, and it is settled. 34 CFR 685.219(c)(3) says a consolidation loan carries the weighted average of the qualifying PSLF payments made on the loans it repaid. We apply that rule, weighted by balance and rounded down to a whole month — no source states a rounding convention, and rounding down never overstates your progress.
What it does not model. Consolidating only some of your loans; which servicer you would land with; and FFEL loans — where a typed FFEL row suppresses the comparison entirely rather than pricing a "before" side under rules those loans do not follow. Consolidation cannot be reversed, so treat this as an estimate to check with your servicer before acting on it.
Which poverty guidelines are applied?
IBR, PAYE, and ICR subtract a multiple of the federal poverty guideline. RepayCompass defaults to the 2026 table, with a visible disclosure under the results and a switch to 2025 under Assumptions.
Why there is a switch at all. Official surfaces do not agree with each other. In July 2026 we saw one StudentAid.gov page quote IBR on the 2025 table and PAYE on the 2026 table — for the same borrower, at the same moment. What separates them is not the calendar: every 2025 sighting also carried a stale family size, which points at a cached record rather than a policy. When you type your own figures into the official calculator — the same thing you do here — it applies 2026, and our numbers reproduce it exactly.
So 2026 is the default. Pick 2025 if a servicer letter or an older official quote used that table and you want to reconcile against it; for a low-income household the two can differ by around 6% on the monthly payment.
How results are tested
The engine runs an automated suite (nearly 300 tests at this writing) including fixtures from a real borrower's official quotes: the StudentAid.gov IDR application's plan-by-plan payments and the Federal Register's ICR worked examples, reproduced within a dollar — most to the cent. Edge cases are tested explicitly: $0 income minimums, very high incomes hitting caps, family sizes past 8, balances that pay off before forgiveness, and the married-filing-separately rules. A property test asserts a RAP balance can never grow month-over-month with on-time payments.
Known conventions and limits
- Rates are blended across your loans. You can enter each loan separately — balance and rate per loan — and for every plan except ICR the engine blends them into a principal-weighted average, which matches how interest actually accrues. Entering one lumped balance with a rate averaged a different way (by outstanding balance, as some dashboards display it) can land a few cents a month off the official Standard quote, so itemizing is the more accurate entry. ICR is the exception: it amortizes each loan at its own rate, because its formula requires it.
- The RAP dependent credit is shown both ways, because the statute includes $50/month per dependent but official tools have been observed quoting without it — pending confirmation with servicers.
- ICR's factor table is held constant across projection years (the 2026 table is the last one published before the plan retires).
- Household proration re-splits every year. When two spouses who file jointly both have loans, each income-driven payment is prorated by that spouse's share of the couple's balances (see "Married borrowers and household math" above). The regulation recomputes the share at each annual recertification, and RepayCompass projects it the same way — your first payment uses today's balances, and from year one on the split follows the balances. Each plan card assumes your spouse is repaying under that same plan, because their balance depends on which plan they are on. The pairing table in the household summary is the exception — each row there is computed for the two plans it names.
- Auto-pay interest reductions aren't modelled. The interest rate you enter is projected as-is for the whole term. Enrolling in automatic payments cuts your rate — temporarily by 1 percentage point instead of the usual 0.25, through June 30, 2028 — and none of that is reflected in these figures, in either direction. See the auto-pay question in the FAQ.
- Parent PLUS isn't supported yet. Those loans follow different rules, so the tool stops rather than showing numbers computed under the wrong ones.
Spot a number that disagrees with an official quote? That's exactly the kind of report we want — send the details (never include your SSN or account numbers). For your actual options, your servicer and StudentAid.gov are the final word.