Why two salary calculators disagree

1. They disagree about what your salary is

"Gross salary" is ambiguous in ways that matter. Does the figure include a fixed allowance, a thirteenth month, a guaranteed bonus, employer contributions, or overtime? Different calculators default differently, and some countries have a statutory definition of "wages" for contribution purposes that excludes items your employment contract calls salary.

This bites hardest where contributions are involved. An allowance that counts as wages for one scheme and not another will change a contribution line without changing anything you would think of as your pay, and a calculator that takes one number cannot know which parts of it qualify.

2. They assume different reliefs and allowances

Almost every income tax system reduces taxable income before applying rates, and almost every calculator has to assume a default. Typically that default is the most conservative one: a single filer, no dependants, no additional deductions, taking only the standard personal allowance.

If you are married, filing jointly where that is permitted, supporting children or parents, paying deductible insurance premiums, repaying a qualifying student loan, or claiming education or medical relief, the default is wrong for you — usually in the direction of understating your take-home. A calculator that asks about marital status and dependants will diverge from one that does not, and the one that asked is closer for you specifically.

3. Table-driven contributions versus percentages

Several countries compute social contributions from statutory tables in wage bands rather than as a flat percentage. Malaysia's EPF is the clearest example: contributions are read from a schedule that steps in wage bands, so the true figure and the percentage approximation differ by small amounts at most salaries.

A calculator that multiplies by the headline rate will be consistently a little off, and it will be off in a way that never quite matches a payslip. Two calculators that disagree by a couple of units on a contribution line, while agreeing on everything else, are almost always a table implementation and a percentage approximation sitting next to each other. The Malaysian breakdown goes through this in detail.

4. Ceilings, floors and thresholds

Most contribution schemes stop at a ceiling: above a certain wage, the contribution is capped. Many also have a floor, and some have a separate threshold at which a different rate applies. These are revised periodically, often at a budget, and a revision instantly makes every calculator that has not been updated wrong for everyone above the old ceiling.

This is the most common cause of a large, sudden disagreement between two calculators for a high salary and no disagreement at all for a low one. If two calculators agree at 3,000 and diverge at 15,000, look for a ceiling one of them has not updated.

5. Which tax year they are using

Tax years do not align with calendar years everywhere, rates change at the boundary, and a calculator may be using last year's bands, this year's, or next year's announced-but-not-yet-effective ones. Around a budget announcement, all three exist simultaneously on different sites.

A calculator that does not state which year it applies to cannot be checked, which is a reasonable basis for preferring one that does. Mid-year rate changes are worse still, because the correct annual figure is then a blend and calculators handle it inconsistently.

6. Rounding and order of operations

Statutory rounding is more specific than it sounds: some contributions round up to the next whole unit, some round to the nearest, some round at each step and some only at the end. Applied across twelve months these differences accumulate into a visible annual gap.

Order matters at least as much. Whether income tax is computed before or after deducting pension contributions changes the answer materially, and the correct order is a matter of statute rather than preference. A calculator that applies tax to gross where the law applies it after a deductible contribution will overstate tax for everyone.

How to tell which one to trust

Three questions settle it. Does it state the tax year or assessment period it applies to? Does it show the bands, tables and rates it used, or only the final number? Does it link the agency that publishes them?

A calculator that shows its working can be checked against the source in a minute. One that returns a single figure cannot be checked at all, and its confidence is not evidence. When two disagree, the one that shows its bands lets you find the discrepancy; the other only lets you pick a side.

Then check the questions it asked. If it never asked whether you have dependants and you do, it is not wrong so much as answering a different question than the one you meant.