Income tax in India, in the order it is actually computed

The calculation is a queue, not a formula

Indian income tax runs in a fixed order: sort every rupee into a head, add the heads into gross total income, subtract whatever deductions your regime allows, apply the slab rates to what is left, subtract any rebate, add surcharge, add cess, then subtract the tax already paid on your behalf. What is left is owed, or refunded.

That sentence is the whole article. Everything below is an explanation of one step in it, and the reason the sequence is worth memorising before any individual rate is that each step takes the previous step's output as its input. A change early does not just alter one line; it moves every line under it. That is why two people can be told the same “tax-saving” fact and get wildly different value from it.

Two places in that queue account for nearly all the confusion. Step 3, because which deductions exist at all depends on which regime you are in — and the default is not the one most people picture. And steps 5 to 7, because almost everyone treats tax as a single percentage of income, when it is a stack of adjustments applied to a number that is itself the output of four earlier steps.

One thing to settle before going further. The Income-tax Act 1961 was repealed on 1 April 2026 and replaced by the Income-tax Act 2025. The rates barely moved. Almost every familiar section number did. Section numbers appear below only where they are the 2025 Act's own.

Step 1 — income is sorted before it is taxed

Tax law does not look at your bank statement and total it. It first asks what kind of income each amount is, because the kind decides everything that follows: which expenses come off it, whether a flat deduction applies, and in some cases whether the income reaches the slab ladder at all.

Take three people who each received ₹60,000 in a year — one as salary, one as rent, one as a gain on selling equity funds. Same money into the same bank account. Three different computations, and potentially three different tax figures on the identical amount.

The row that does the most quiet work is the last one. “Other sources” is a catch-all, which is why “it wasn't salary” has never been an argument for anything. Interest on a savings balance, interest on a fixed deposit, a maturity payout that is not exempt — the list is not closed because the head is defined by exclusion. Income does not escape by failing to fit one of the named categories; it falls into the one designed to catch it.

Notice also the house-property line, because it is the clearest example of a head-specific rule that has nothing to do with what you actually spent. A fixed proportion of the annual value comes off as a deduction regardless. A landlord who maintained the flat beautifully and a landlord who did nothing get the same allowance. The head, not the behaviour, decides.

Step 2 — gross total income, and what set-off does

Adding the heads together sounds like the least interesting step. It is where losses are dealt with, and losses are worth real money if they are handled correctly and worth nothing at all if they are not.

The principle is that a loss under one head can be set against income — sometimes under the same head, sometimes across heads, and the rules differ by the class of loss. What cannot be used this year can generally be carried forward and set against future income of the permitted kind, for a limited number of years.

Here is the mechanism that makes this a personal-finance point and not a compliance footnote. A capital loss is not a sunk cost; it is a credit against a future tax bill. Selling a holding at a loss in a year when you also booked a gain reduces the gain that gets taxed. Selling it in a year with no gain does not waste it either, provided the loss is carried forward properly.

And that proviso is where people lose money for paperwork reasons. Carrying a loss forward has historically depended on the return being filed by the due date — a condition worth confirming for the year you are filing, because it converts a filing formality into a money decision. Someone with no tax payable, who therefore assumes there is nothing to file, can forfeit a carry-forward they would have used two years later.

The output of this step has a name worth keeping straight, because the next step changes it: this is gross total income. It is not the figure the slabs are applied to.

Step 3 — deductions, and the fork that decides most of the outcome

This is the step where the majority of what people believe about Indian income tax is now out of date, in two separate ways at once.

First: the new regime is the default. Under s.202 of the Income-tax Act 2025, a taxpayer is in the new regime unless they actively opt out of it. Not choosing is a choice, and the thing it chooses is the regime in which the familiar deductions do not exist.

Second: the section numbers moved. The 1961 Act was repealed on 1 April 2026. The deduction everyone still calls 80C now lives at s.123 read with Schedule XV. Health insurance premium is s.126. Education-loan interest is s.129. The rates changed hardly at all; the citations changed completely.

Read the right-hand column of the first four rows together, because that is the whole design. The new regime buys lower slab rates with the deductions. A short list of allowances survives it; the large, familiar ones do not. Presenting the investment deduction as something every taxpayer can claim describes a minority position, not the default one.

Which regime produces less tax for a given person is arithmetic, not opinion, and it turns on a single quantity: the total deductions they would actually claim. Below some amount of deductions the lower rates win; above it the deductions win. That crossover point is computable from your own numbers by running the calculation twice, once each way, and it is not the same for two people with the same salary. Nobody can tell you which side of it you are on without your figures — the comparison itself is worked through in the old regime against the new.

The other half of the trade-off is rarely stated. A deduction under s.123 — available, remember, only to someone who has opted out of the default regime — is not a discount on a purchase you were making anyway. It is a discount on money you have agreed to lock up, often for years. The tax saved arrives once. The lock-in lasts the full term. That is a real cost, and it belongs in the comparison next to the saving, particularly for money that has a job to do before the lock-in ends — see the order of operations for where a tax-driven commitment sits relative to everything else, and tax-saving investments for what the eligible instruments actually are once the deduction is set aside.

One transition note, because both statutes are live at the moment. Income for FY 2025-26 — the financial year that ran from April 2025 to March 2026 — is still governed by the repealed 1961 Act, so a return being filed for that year carries the old section numbers and they are the correct ones on it. Income from 1 April 2026 falls under the 2025 Act. Two returns filed in the same calendar year can legitimately cite different statutes.

Step 4 — the slab ladder taxes slices, not people

The single most common misunderstanding in Indian personal finance is that a slab rate applies to your whole income. It does not, and the arithmetic of why is short enough to do here.

Suppose — and these three numbers are invented in this paragraph to make the shape visible, not taken from any slab table — that nothing is charged on the first ₹10 of income, 10% on income between ₹10 and ₹20, and 30% above ₹20. Someone earning ₹22 pays nothing on the first ₹10, one rupee on the next ₹10, and 60 paise on the last ₹2. Total: ₹1.60, which is 7.3% of what they earned, even though the top band they touched charges 30%.

Two rates are in that example and they answer different questions. The 30% is the marginal rate — what the next rupee of income costs. The 7.3% is the average rate — what the whole income cost. Conflating them is what produces the belief that a raise can leave you worse off, and under a pure slab ladder that is mechanically impossible: the higher rate only ever touches the rupees above the line, never the ones below it.

The real bands are published each year and change. The shape does not, which is why the shape is the thing worth carrying around.

Now the honest qualification, because the folk belief is not baseless — it is attached to the wrong step. There are points in the Indian calculation where an extra rupee of income does cost more than a rupee. They are not in the slab ladder. They are two steps further down, and they are the subject of the next section.

One more thing this step does: it takes certain income out. Capital gains taxed at their own flat rates are lifted out of the ladder and charged separately, then added back into the total tax. So a taxpayer's slab position and their tax on a share sale are largely independent facts — covered below.

Steps 5 to 7 — rebate, surcharge, cess, and why the order is not cosmetic

Three adjustments sit between the slab computation and the final liability. Each one operates on the output of the one before, which makes their sequence part of the answer rather than a presentational detail.

The rebate reduces the tax, not the income. It applies where total income is at or below a stated level and it does nothing at all above that level. This is the reason a taxpayer can have a positive tax computed at step 4 and owe nothing at step 5.

Surcharge is an add-on percentage of the tax, switched on once total income crosses a stated threshold and stepping up again at higher ones. The thresholds form a ladder; each individual step is a cliff rather than a slope. Cross one of those lines by a rupee and the surcharge applies to the whole tax, not merely to the rupee that crossed. Left unpatched, that would mean an extra rupee of income costing more than a rupee in tax — the one outcome a tax system cannot defend. So the Act patches it: marginal relief caps the additional tax at the additional income. The existence of the patch is the proof that the cliff is real.

Cess is charged on the tax, not on the income. Its size therefore depends entirely on what steps 5 and 6 produced. A taxpayer whose liability falls to nil at the rebate pays no cess either, because a percentage of nothing is nothing. Move the cess ahead of the rebate in the queue and the same person would owe something. The order is doing arithmetic work, not just organising a page.

Why a deduction is not worth what you think it is worth

Ask most people what a deduction is worth and they will say: the amount, multiplied by your slab rate. Claim a deduction while in the 30% band and you save 30 paise in the rupee. That is a straight line, and across most of the range it is correct.

The rebate puts a step in that line. Below the threshold the tax is wiped out; above it, the rebate contributes nothing. So for a taxpayer sitting a little above the threshold, a deduction that pulls total income under it does not save its amount times a rate — it saves the entire tax bill. The identical rupee of deduction is worth a few paise at one point on the curve and hundreds of times that a few thousand rupees along it.

The corollary runs the other way at the surcharge threshold, where an extra rupee of income costs disproportionately until marginal relief intervenes. A tax bill is not a smooth function of income. It has steps in it, and the steps are where planning either works unusually well or backfires. Where those steps sit changes with every Budget. That they exist does not, and knowing they exist is what tells you when a rule of thumb has stopped applying.

This is a mechanism, not a strategy. It says where to look, not what to do, and it says nothing at all about whether committing money to a deduction-eligible instrument is a sensible use of that money — a question the tax saving cannot answer on its own.

Which income actually escapes the slab ladder

Some income is taken out of the slab computation entirely and charged at its own flat rate. Capital gains are the case almost every household meets — though, as below, not all of them escape.

For equity shares and equity-oriented fund units on which securities transaction tax was paid, the 2025 Act sets two rates. Gains on units held for up to 12 months are short-term and taxed under s.196 at 20%. Beyond that they are long-term, and s.198 taxes them at 12.5% on the amount above ₹1.25 lakh of such gains in the year. A fund counts as equity-oriented when more than 65% of its proceeds sit in domestic equity shares — the test is the portfolio, not the name on the scheme.

Units of a specified mutual fund — broadly one whose portfolio sits predominantly in debt and money-market instruments, the exact proportion being a stated statutory test — acquired on or after 1 April 2023 are dealt with by s.76 instead, and taxed at slab rates, with the gain always treated as short-term. There is no holding-period benefit to wait for.

Read that last one carefully, because it is the exception that proves the sorting. An s.76 gain does not get a flat rate of its own; it is sent back to the slab ladder and charged like any other income at the top of your stack. Being “capital gains” is not by itself what takes income out of the ladder. Being a named class of asset held for a named period is. A gain outside those named classes is computed under this head and then taxed with everything else.

Two consequences follow from a flat rate, where one applies, that are worth stating plainly, because they cut in opposite directions.

A flat rate is indifferent to the rest of your income. Two people with very different salaries pay the same tax on the same equity gain. For a high earner that is favourable relative to their slab rate; for someone with modest income it is the reverse, and it is why “my income is below the exemption limit so I owe nothing” can be wrong when the income in question is a capital gain rather than salary. Whether an unused basic exemption can be set against such gains is a specific question worth confirming for your year rather than assuming in either direction.

The tax lands when you realise the gain, not as it accrues. That is the deferral mechanism, and it is the single most underrated force in long-horizon investing: money that has not yet gone to tax is money still compounding for you. A deposit taxed on accrual every year loses the compounding on the tax as well as the tax; a holding taxed only on sale does not. Over a year the difference is negligible. Over fifteen it is doing serious work — the arithmetic is set out in the time value of money, and the comparison it drives is worked through in debt funds against fixed deposits.

What the rate does not tell you is anything about the investment. A holding sold purely to avoid crossing a gains threshold has had its investment decision made by the tax code, which is a poor place to make it. Where the equity-oriented test actually bites in practice is set out in equity funds explained, and the head itself — cost of acquisition, holding periods across asset classes, and how a gain is computed rather than merely rated — is covered in capital gains tax.

Past returns on any holding describe what happened, not what will. Nothing here forecasts either.

Step 8 — tax already paid, and why a refund is not a windfall

By this point the liability is known. The last step compares it against what has already reached the government on your behalf during the year, and the gap is settled in one direction or the other.

Three things arrive in this step.

TDS — tax deducted at source. A payer withholds tax from what it pays you and deposits it against your permanent account number — your PAN. Here is the mechanism people miss: the deductor can only see its own payment to you. Your employer does not know about your freelance income, your bank does not know your regime, and neither knows what deductions you will claim. TDS is a third party's estimate of a tax that only your return can compute. It is right by coincidence, not by design, and it is wrong in both directions routinely.

Advance tax. The Act wants tax paid as income arises rather than in a lump at the end, so it requires instalments during the year and charges interest where they fall short. Salaried taxpayers usually meet this through TDS without noticing. Anyone with capital gains, freelance receipts or substantial interest income meets it the hard way, because no deductor is withholding enough on those — and gains in particular are invisible to every deductor involved.

Self-assessment tax. Whatever is still short when the return is prepared, paid before filing.

Which brings us to the refund. A refund is not money earned; it is the return of an overpayment. A large refund every year means a large sum sat with the government through the year instead of with you — which is not a disaster, but it is not a win either, and it is worth recognising as a cashflow outcome rather than a bonus. Whether the department pays interest on it, and from when, is a separate question worth checking.

One practical point that has changed the shape of filing. The department already holds a statement of what banks, employers and registrars reported against your PAN — the annual information statement, and the tax credit statement alongside it, long known as the AIS and Form 26AS, though the names and formats are worth confirming for the year you are filing. Reconciling your return against those before filing rather than after a notice is most of the work, because mismatches are detected mechanically and at scale. An interest entry you forgot is not hidden; it is already on the department's copy.

Seven errors that come from skipping the sequence

Each of these is specific, common, and traceable to a particular step above.

1. Treating TDS as the tax. Step 8 confused with step 7. The deductor computed a withholding on one slice of your income in isolation; nobody computed your liability but you.
2. Assuming the old-regime deductions apply by default. Step 3. The new regime applies unless you opt out, and under it s.123, s.126 and s.129 are not available.
3. Reading the investment deduction as a tax saving of the same size. A deduction of ₹1.5 lakh under s.123, which exists only for someone who has opted out of the default regime, reduces the income that is taxed by that amount, not the tax by that amount. The saving is that figure multiplied by the rate that would otherwise have applied — except near a rebate threshold, where it can be much larger.
4. Fearing a raise because of a bracket. Right worry, wrong step. The slab ladder cannot make you worse off; the rebate and surcharge thresholds can, which is why marginal relief exists.
5. Not filing in a loss year. Step 2. No tax payable is not the same as nothing to file, and a carry-forward forfeited now is a deduction unavailable against a gain two years out.
6. Assuming income below the exemption limit means no tax. True of slab income. Not automatically true when the income is a capital gain charged at a flat rate outside the ladder.
7. Citing repealed section numbers. The 1961 Act governs FY 2025-26 and nothing after it. Planning done today against 80C rather than s.123 is planning against a statute that no longer exists. In conversation that is harmless shorthand. It stops being harmless the moment a provision's substance moved along with its number, and the only way to know whether it did is to read the 2025 section rather than recall the 1961 one.

None of the seven is exotic. All seven are sequence failures rather than knowledge failures, which is the argument for learning the queue before learning any rate.

Which parts of this are arithmetic

Most of the sequence above needs facts rather than judgement, and two of those facts are the ones people reconstruct badly from memory at filing time: when a holding was bought, and what it was worth on each date in between.

The holding-period line at 12 months decides whether s.196 or s.198 applies to a fund sale, and the answer turns on acquisition dates rather than on anything the tax code can be argued about. FNOTrader's Mutual Funds app runs on the full published history of daily per-unit scheme values from AMFI — the net asset value, or NAV — around 34 million rows of it. For any scheme and period it reports invested amount against value, maximum drawdown, and the annualised return on cashflows that land on irregular dates, which is XIRR. That is the transaction picture a gain computation takes as its input.

It does not compute anyone's tax, and nothing on this page is tax advice. FNOTrader is not a chartered accountant, a tax practitioner or a SEBI-registered investment adviser. What the article is for is making the shape of the calculation visible, so that the figures — which change every February — land somewhere.

Common questions

Continue reading

More in Tax Planning · App: Mutual Funds · Definitions: glossary · Free tools: calculators · All: every article