About the student loan engine

What this is

A decision engine, not a payment calculator. You enter your real loan mix; the engine simulates all nine federal repayment plans month by month over up to 30 years and returns a ranked recommendation by total lifetime cost — with the irreversible choices flagged before you make them. The cheapest monthly payment is frequently the most expensive plan, because a smaller payment stretches repayment across more years of interest. Showing that conflict — and the plan switches that cannot be undone — is the point.

Why can't a single-plan calculator answer this?

Because the plans interact, and the cheapest month is not the cheapest decade. RAP has no payment cap: 34 C.F.R. § 685.209(f)(5) simply omits the “lesser of” clause that paragraphs (f)(2) and (f)(3) carry for IBR and PAYE, so a high income with a moderate balance can make RAP cost more than the 10-year Standard plan. A calculator that only knows one plan cannot show you that crossover, and neither can one that stops at the monthly figure.

Three further interactions decide real cases and only appear when every plan is run side by side: forgiveness outside PSLF is taxable income under current federal law, so a plan that forgives more is not automatically cheaper; PAYE and ICR end on 1 Jul 2028, so any projection past that date has to model the forced move rather than quietly project PAYE for 20 years; and switching to RAP forfeits qualifying payments already made toward IBR, PAYE or ICR forgiveness. The engine ranks the plans on total lifetime cost, and marks that forfeiture as the one irreversible result it reports.

Which rules does it encode?

Every rate, bracket, term and threshold comes from a versioned rule file citing its primary source. The governing document is the RISE final rule — 91 Fed. Reg. 23768, published 1 May 2026 (FR Doc. 2026-08556), effective 1 Jul 2026 — which amends 34 C.F.R. parts 674, 682 and 685.

• RAP: eleven AGI bands from § 685.209(b)(2), running from a flat $120 annual base at or below $10,000 of AGI to 10% above $100,000; a $50 monthly reduction per dependent; a $10 monthly floor; the interest waiver at § 685.209(h)(4)(i); the $50 principal match; forgiveness at 360 payments; PSLF at 120.
• The income-driven plans: Old IBR at 15% of discretionary income with forgiveness at 300 payments; New IBR and PAYE at 10% with forgiveness at 240; ICR at 20% measured against 100% of the poverty guideline rather than 150%, with a 12-year alternative amortisation.
• The fixed plans: Standard over 120 months; Tiered Standard over 120, 180, 240 or 300 months by balance; Graduated over 120 months with no payment more than 3× any other; Extended over 300 months, available above $30,000.
• The poverty guidelines: the 2026 HHS table — $15,960 for the first person in the contiguous 48 states and DC plus $5,680 for each additional person, with separate figures for Alaska ($19,950 / $7,100) and Hawaii ($18,360 / $6,530).

What does it deliberately not model?

State tax treatment of forgiven balances is out of scope in v1, and the rule file says so rather than assuming a rate. Your loans are aggregated into one balance at a balance-weighted average rate, where a servicer amortises each loan separately. A plan is shown as eligible only when every loan you entered can use it, though real borrowers can sometimes hold different plans on different loans. Poverty guidelines are held flat across the projection, which is the conservative direction — a rising guideline would lower income-driven payments over time. The full list, with the reasoning for each, is on the methodology page.

Which figures are still unverified, and against what?

Six, listed below with the source that would close each one. Nothing has been filled in with an estimate. The same register, summarised across all five tools, is on about Bracketsight.

• The assumed marginal tax rate of 22% used to estimate tax on forgiveness is a flat modelling estimate, not a regulatory figure. The correct rate is borrower-specific, because the forgiven amount stacks on top of other income.
• ICR's income-percentage factor is fixed at 1.0. 34 C.F.R. § 685.209(f)(4)(i)(A) multiplies the 12-year amortisation by a factor the Secretary sets in a Federal Register notice published annually, and the 2026 notice has not been located. Until it is, ICR is understated for higher incomes.
• Graduated repayment's 24-month step interval has no regulatory basis at all. § 685.208(b)(6)(i) prescribes only “payments at two or more levels”; the interval is a servicer convention. The 10-year term and the 3× ratio around it are verified.
• The taxability of non-PSLF forgiveness rests on the 2024 U.S. Code edition of 26 U.S.C. § 108(f), because no current official edition could be retrieved. A 2025–26 amendment cannot be ruled out from a primary source.
• Three citations point at congress.gov, which refuses automated requests, so none of the three has been read. Every figure they would support is independently verified against the RISE rule text.
• The new-borrower tests for PAYE and IBR each model one limb of a two-limb rule. The engine can therefore offer PAYE to someone who held a balance on 1 Oct 2007, and can apply New IBR's terms to someone who has taken a loan on or after 1 Jul 2026. Both need an input field the form does not yet collect.

Has a credentialed professional reviewed this?

No. No student loan professional has been engaged, and no review has taken place. The engine is unit-tested, with eight worked RAP examples held as golden cases, and every rule file records the primary source it was read from and the date it was last checked — but confirming that an encoded value matches its cited source is a much narrower question than whether the output is sound for your situation.

How should you use it?

Use it to narrow nine plans to two or three, and to find the questions worth asking. The ranking is arithmetic over published rules, so it is reproducible: every figure opens its own formula, inputs and rule version. Take the shortlist and the flagged irreversible steps to your servicer and confirm them against your actual account before you elect anything.

Do not use it as a record of your loans. It does not connect to StudentAid.gov or to any servicer, cannot see your payment history, and knows only what you typed. If your balances, rates or disbursement dates are approximate, so is everything downstream of them.

What this is not

Not your servicer, not financial advice, not tax advice. Estimates run under current rules and your stated assumptions. Confirm any plan change with your servicer, and any tax question with a professional, before acting. Found an error? Write to info@bracketsight.com — the editorial policy sets out what happens next, and the changelog records every rule change that follows.