Exactly which official source backs every number on this site, and what's still honestly incomplete.
Every calculator on this site is checked against a primary, official source before it goes live — not a secondary blog post or aggregator site. Where we found genuine disagreement between sources during research, or where full coverage isn't there yet, that's disclosed directly below rather than hidden. This page exists so that claim can actually be verified, not just taken on trust.
Tax year 2026 brackets and standard deductions from IRS Revenue Procedure 2025-32, the same document the IRS itself publishes each year to set the following year's inflation-adjusted figures. Verified against multiple independently published worked examples before use.
IRS.gov source →All 9 no-income-tax states and all genuinely flat-rate states are fully verified. Graduated-bracket states were sourced from the Tax Foundation's annual report, cross-checked against each state's own published cumulative-tax formula where available (for example, California and New York were verified to the cent against their own tax authorities' worked examples). Married-filing-jointly brackets are now modeled for every state that publishes different thresholds for joint filers (Alabama, Connecticut, Hawaii, Kansas, Maine, Montana, Nebraska, New Mexico, Oklahoma, Oregon and Vermont, alongside California, Maryland, Minnesota, New Jersey, New York and Wisconsin, which already had them). Nine further graduated states — Arkansas, Delaware, Massachusetts, Missouri, Rhode Island, South Carolina, Virginia, West Virginia and the District of Columbia — apply one schedule to every filing status, and the calculator now says so explicitly rather than hedging. Head-of-household thresholds are still approximated with single-filer thresholds in the eleven states listed first, which is disclosed in the calculator's own result whenever that filing status is selected.
2026 rates, wage base, and Additional Medicare Tax threshold cross-referenced against multiple independent payroll and CPA-published sources before use.
2026 federal brackets and Basic Personal Amount, including the BPA phase-out range for higher incomes, sourced directly from the Canada Revenue Agency's own published figures.
Canada.ca source →12 of Canada's provinces and territories use official 2026 bracket rates published directly by the CRA. Two honest gaps: these figures don't yet include each province's own basic-personal-amount-equivalent credit (meaning real amounts owed are likely somewhat lower, especially at lower incomes), and Ontario and Prince Edward Island's additional provincial surtax isn't included.
Quebec is administered separately by Revenu Québec on a genuinely different tax base, not just different numbers — it required its own dedicated sourcing. Both the 2026 bracket rates and the $18,952 Basic Personal Amount credit are sourced directly from Revenu Québec's own published rates page and the Quebec Ministry of Finance's official 2026 indexation parameters, cross-checked against the source's own worked example before use. The federal 16.5% Quebec abatement is also correctly applied to the federal estimate when Quebec is selected.
Revenu Québec source →2026 contribution rates and maximum pensionable/insurable earnings cross-referenced against multiple independent Canadian payroll sources before use.
These calculations use the same standard formulas taught in financial mathematics and used industry-wide by lenders and financial institutions — the fixed-payment amortization formula, the compound interest formula with periodic contributions, and their mathematical inverses (solving for time-to-payoff or required contribution instead of payment). These aren't sourced from a single external site; they're verified through internal consistency checks — for example, confirming that solving for a payoff time and then feeding that time back into the original payment formula returns the original payment, to the cent.
Both new tools run the loan month by month rather than using a closed-form shortcut, because that is what a servicer does: interest accrues on the balance actually outstanding, and the final payment is trimmed to whatever remains. The extra-payment tool simulates the baseline loan the same way it simulates the accelerated one, so a comparison between them is like-for-like and an extra payment of zero correctly reports a saving of zero rather than a few dollars of rounding artefact.
The refinance tool reports two figures that frequently disagree: the break-even month, which is closing costs divided by the monthly saving, and the lifetime cost difference, which includes the effect of resetting the loan term. A refinance can break even quickly and still cost more overall, and the calculator says so explicitly when that is the case rather than leading with the flattering number. Neither tool models tax treatment of interest, mortgage insurance, cash-out amounts, or prepayment penalties.
This is the only tool on the site that cannot produce a single correct answer, and it is built to say so. Both paths are compared on net worth at the end of the period: the buyer holds the home's value less selling costs and any remaining mortgage, the renter holds a portfolio seeded with the down payment and closing costs the buyer spent. Whoever has the lower monthly outgoing invests the difference — in both directions, since a model that only ever lets the renter invest is tilted toward buying without admitting it. Property tax and maintenance are charged against the home's current value rather than its purchase price, and PMI applies while the loan exceeds 80% of the original price.
Every assumption is an input rather than a hidden constant, and the tool reports the home appreciation rate at which the two options come out level, because that figure is checkable against what a reader knows about their own market in a way a verdict is not. The mortgage interest deduction is deliberately excluded, since most filers now take the standard deduction and including it by default would overstate the case for buying for the majority of users. Capital gains treatment, moving costs, and the non-financial value of either option are also outside the model.
The sales tax tool applies a rate you supply rather than a stored table, because combined US rates vary by exact address within a state and a state-level table would be wrong for most transactions. The inflation tool is rate-based for the same reason: a published price index embeds a choice of index, basket and base year, and the assumption belongs to the reader. The interest rate tool solves the amortization formula numerically by bisection, since that formula cannot be rearranged to isolate the rate; precision is limited by the fact that the payment you enter is rounded to cents.
The amortization schedule computes each month’s interest as the outstanding balance times the monthly rate, rounded to whole cents, with the final payment trimmed to clear the balance exactly — the method a servicer uses. This differs from the closed-form total (payment times term, less principal) by roughly $4.71 over a 30-year loan, under a cent per payment. Both figures are returned so the difference is visible rather than surprising. The investment tool reports nominal and inflation-adjusted balances together, and the retirement tool inflates spending to the retirement date before drawing down. Neither models variable returns, and the retirement tool excludes Social Security, pensions, and tax on withdrawals, all disclosed on the page itself.
Conversion factors are held in a separate module from the financial logic, so a fault in one cannot disable the other. Almost every factor used is a definition rather than a measurement, and therefore exact: the 1959 International Yard and Pound Agreement fixed the inch at exactly 2.54 cm and the pound at exactly 0.45359237 kg, and every other imperial length and weight follows from those two. The US gallon is exactly 3.785411784 litres by its definition as 231 cubic inches; the Imperial gallon is exactly 4.54609 litres under the UK Weights and Measures Act 1985. A nautical mile is exactly 1,852 metres, a hectare exactly 10,000 square metres, a bar exactly 100,000 pascals, and the thermochemical calorie exactly 4.184 joules.
Two conversions are not simple multiplications and are handled separately. Temperature requires an offset as well as a scale factor, since the two scales have different zero points. Fuel economy is an inverse relationship, because miles per gallon measures distance per volume while litres per 100 km measures volume per distance — doubling one halves the other.
Two are genuinely approximate, and say so on the page. Cups to grams depends on the density of the specific ingredient, and even then varies with how the ingredient is packed: flour measured by scooping can weigh twenty percent more than flour spooned into the cup. Digital storage has two competing definitions in active use, decimal and binary, and the tool asks which you mean rather than picking one silently.
The mortgage calculator now runs the loan month by month rather than holding every cost static. Property tax, insurance and HOA dues escalate by their own annual rates if you set them; PMI is charged while the balance exceeds 78% of the original purchase price and stops automatically at that point, which is the automatic-termination threshold under the US Homeowners Protection Act. Extra payments — monthly, annual, or a single lump sum — are applied to principal and shorten the loan rather than reducing the required payment. Payment numbers are converted to real calendar dates from the start date you enter, so the result is a payoff month rather than a payment count.
The biweekly comparison models half the monthly payment paid every fortnight, which is twenty-six half-payments a year and therefore one extra monthly payment annually. Interest there accrues on a weekly basis rather than monthly, which is why its total differs slightly from simply adding one payment a year to the monthly schedule. The comparison assumes a servicer that applies each half-payment on receipt; some hold the money and disburse monthly, which removes the benefit entirely, and the calculator says so on the page.
Every chart on the site is inline SVG generated in the browser from the same engine that produces the numbers beside it, so a chart can never disagree with the figure it illustrates. No charting library is used: the whole module is about six kilobytes, it prints without a separate stylesheet, and it inherits the site’s colours rather than hard-coding them.
Each chart carries a text description of what it shows, because a chart that conveys nothing to a screen reader takes up space without adding meaning. Where a visual would be decorative rather than informative — the number-to-words converters, for instance, where there is no quantity to compare — none is drawn.
Ten calculators export their schedule as a spreadsheet file. Rows are rebuilt from the calculation engine at the moment the button is pressed rather than read from the table on screen, so the exported figures carry full precision and always match the current inputs rather than whatever was displayed when the page loaded. Each file opens with a short header recording the inputs it was generated from, so a schedule handed to a lender or an accountant carries its own assumptions with it.
The file is assembled entirely in your browser and never sent anywhere. A byte-order mark is written at the start so spreadsheet software reads it as UTF-8 rather than mangling currency symbols, and any field containing a comma, quote or line break is quoted with internal quotes doubled.
Four rules govern every number on this site.
Primary sources only. Figures come from the body that sets them — IRS publications and Revenue Procedures, CRA guidance, and each state or province's own revenue authority or statute. Where only secondary sources exist, they are cross-checked against one another and the limitation is disclosed in the calculator's own result rather than buried here.
Disagreement is reported, not resolved by guessing. When credible sources conflict, the conflict is stated and neither side is presented as settled. Pending legislation is not modelled as if it were law; where a bill would change a figure, it is checked for enactment before anything moves.
Math is verified by hand, not by inspection. Calculations are checked by recomputing a worked example independently and comparing it to the code — not by reading the code and judging it correct. Changes are tested through the rendered page, and the remaining calculators are re-run afterwards to catch anything the change broke elsewhere.
Errors are published. Every correction is recorded on the Corrections & Changes page, dated, with the size of the error stated in dollars wherever it can be quantified. Approximations are labelled as approximations inside the tool that produces them, so a caveat can never be missed by someone who never scrolled this far.
MyLoanMath is free and supported by advertising. No calculation, recommendation, or tool on this site is influenced by that: there are no affiliate links to lenders, banks, insurers or tax preparers anywhere, and no result is shaped by anything commercial. Every tool runs entirely in your browser and nothing you type is transmitted anywhere — see the privacy policy for detail. Corrections are welcome and genuinely wanted at contact@myloanmath.com.
Tax figures change annually. The IRS typically publishes the following year's figures in October or November; the CRA follows a similar autumn schedule. We review and update the tax calculators against newly published figures each year once they're officially released, and every tax page shows a "last verified" date so you can see at a glance how current the numbers are.