Sunk Cost vs Opportunity Cost: Which One Belongs in a Decision

A sunk cost is money that has already left and cannot come back whatever is decided next, so it is excluded from the decision. An opportunity cost is what the money or the asset could be earning in its best alternative use, so it is included even though nobody pays it. One test separates them: does this rupee change if the decision changes?

Everything below rests on what a decision actually is. A decision compares two futures, and only the things that differ between those futures belong in the comparison. Put that one sentence in front of any cost and the sorting becomes mechanical. A sunk cost is identical in every future being chosen between, so including it cannot change which option wins and can only bend the reasoning around it. The resource is used here in one future and somewhere else in the other, so an opportunity cost differs between futures by construction. Accounting records payments, which is why it captures the first perfectly and the second not at all. The mismatch between what accounting records and what a decision needs is where nearly every error in this area starts.

Sankalp Industrial Systems Limited, an invented manufacturer, and Sthira Capital Partners, the invented firm that buys it, supply every figure below. The arithmetic settles which costs enter a decision. Whether a buyback, a buyout or a change in a cash balance is worth undertaking is a separate question of judgement.

What makes a cost sunk, if it is not age and it is not size?

The wrong answers are the ones people reach for, so they come first. A cost is not sunk because it is old: last month's electricity bill and last decade's factory purchase are both historical, and neither fact settles anything. A cost is not sunk because it is large. A big number feels harder to ignore, and that is a fact about the person looking at it rather than about the number. A cost is not sunk because it hurt.

A cost is sunk when the rupee is the same in every future being chosen between, so no decision can bring it back or send it away. That is the whole definition and it has a clean consequence: sunkness is not a property the rupee carries around with it, it is a property of the rupee against a particular decision. Ask about a different decision and the same rupee can be live again.

A deposit paid for a wedding hall makes the point. While the hall is still holding the booking and the deposit is refundable, the money is genuinely on the table: cancelling returns it, proceeding does not. The deposit sits inside the choice. The day the refund window closes, the same rupees stop being part of the choice altogether. The wedding can go ahead or be called off and the deposit is gone either way. Nothing about the amount changed. The deposit simply stopped being able to differ.

Notice the two claims the drawing withholds: that the fee stopped mattering, and that nobody need think about it again. Neither is on it. The claim the drawing does make is narrower and more useful. The fee lost its ability to separate one option from another, and separating options is the only job a number has inside a decision.

What is an opportunity cost, when nobody ever paid it?

Here is the harder half, and it is harder for a structural reason: there is no receipt. An opportunity cost is the value of the best alternative use of a resource, given up because the resource is being used here instead. The decision-maker is worse off by that amount than they would otherwise have been, so an opportunity cost is real in the only sense that matters. The record of what was paid is the only place people usually look, and the cost never appears there.

The everyday version sits in most households. Suppose a house has a spare room and a cousin stays in it rent free. Ask what the room costs and the honest answer is not nothing. If the room could be let for a certain amount a month, that amount is what the arrangement costs, whether or not anybody ever discusses it. Nobody writes a cheque. No bank statement moves. The household is still poorer by the rent it is not collecting, and if it ever compares keeping the room free against letting it, the forgone rent is the entire substance of the comparison.

An opportunity cost is what was given up to have what was chosen, and it is included precisely because it differs when the choice changes. Use the room for the cousin and the rent is forgone; let the room and the rent arrives. Two futures, one number that moves between them. The test is the same one as before, running the other way.

The resource does not have to be money, and a second everyday version shows why. A street vendor has one cart and one pitch outside a college gate. Selling snacks there means not selling anything else there, and the best of the things not sold is the cost of the thing sold. The cart is the resource, the pitch is the constraint, and the alternative menu is the price. A vendor who thinks the cart is free because it is already bought has confused an item that is paid for with an item that is unconstrained.

The one test, in one line

Both sides are now defined, so the contrast can be drawn honestly. The two costs differ on exactly one axis, and every other difference between them follows from that one. Ask of any rupee: if I decided the other way, would this rupee be different? If no, take it out. If yes, put it in, even if nobody is paying it.

Run the test on four costs in a row and it stops needing thought. Take the four below, all of them attached to Sankalp Industrial Systems Limited and all of them locked figures from its record.

Why is accounting complete on one of these and silent on the other?

A reasonable reader asks at this point why, if an opportunity cost is real, the accounts do not carry it. The answer is not that accountants overlooked it. Accounting is built to record what was paid and received, so it captures a sunk cost perfectly and an opportunity cost not at all, and that is what it is for rather than a defect in it.

A payment was made and a payment leaves a trace, so every rupee of the Rs 52,00,00,000 of fees appears somewhere in the books of Sthira Capital Partners. Nothing was ever paid for the 12.00 per cent charge for capital, so not one rupee of it appears anywhere in the accounts of Sankalp Industrial Systems Limited. No supplier invoiced it, no bank debited it, no auditor confirmed it. The charge measures the return identical money would fetch somewhere else carrying identical risk, and a ledgerThe running book of what a business actually paid out and took in, entry by entry, with a counterparty behind each line. has no line for a return that somebody else earned.

The sunk instance: Rs 52,00,00,000 that bought nothing

Now the cleanest demonstration in the whole record, and it needs no argument at all because it is a subtraction. Sthira Capital Partners buys Sankalp Industrial Systems Limited. Two of the numbered uses of funds are fees, and neither of them buys an asset.

Watch what happens on the day of completionThe moment a purchase legally finishes and the money actually moves, as distinct from the day terms were agreed.. Value the whole business, take off the borrowings raised against it, and what is left is the stakeThe slice of a company's equity that one holder has a claim on, after everyone with a prior claim has been satisfied. the sponsorIn a purchase of this kind, the party that puts up the equity cheque and runs the business afterwards. How such a purchase is structured is covered separately. holds by that evening.

Nothing went wrong between the morning and the evening. No asset lost value, no forecast was missed, no market moved. The shortfall is the two fee lines and there is nothing else in it.

The fees are sunk the moment they are paid. From the second after completion, no decision the sponsor makes can recover a rupee of them. Sell tomorrow, hold for five years, refinance the borrowings, do nothing at all: the Rs 52,00,00,000 is identical in every one of those futures, so it belongs in none of the comparisons between them. Accepting that requires no belief about the business, only the observation that the number does not move.

A record and an input are different jobs

And yet the figure does not disappear from the sponsor's paperwork, and the confusion starts there. Take the eventual result apart into what produced it and one of the four lines is the fees, minus Rs 52,00,00,000. Those fees come to minus 3.09 per cent of everything the holding period created. The four lines reconcile to the rupee, and the fee line is one of them.

Each share in that last column was rounded once, straight from its full value to the two places printed, so the column lands on 100.00 with nothing nudged. Round the fee share twice instead, to four places and then to two, and it prints as minus 3.10 and the column stops adding up. A figure can be honest history and irrelevant arithmetic at the same time, and knowing which of the two jobs is being done is the whole skill. A decomposition explains a result that already happened. A decision compares futures that have not happened. The fee line is complete and fair in the first job and has nothing to contribute to the second.

The opportunity instance: Rs 80,00,00,000 that is not free

Now the other side, on the same company. Sankalp Industrial Systems Limited holds Rs 1,20,00,00,000 of cash. Of that, Rs 40,00,00,000 is what running the operation from one week to the next requires in hand, and Rs 80,00,00,000 sits over and above that requirement. The split is part of the company's own record rather than an estimate.

Ask a room of people what the surplus costs the company. Most will say nothing, on the ground that the company already has it. Saying nothing is the error in its most respectable form. Money the company already holds has a cost equal to whatever it could earn in its best alternative use, so it is not free, and that cost is borne whether or not anybody writes it down. The surplus is sitting somewhere earning something. Spend it and that something stops. Nothing about ownership changes the arithmetic.

Two things follow immediately. First, the operating Rs 40,00,00,000 is a different animal: it is not genuinely available for any other use, so there is no alternative to give up and nothing to charge. Where exactly a company draws that line is a judgement it makes, and how that judgement is made is covered separately. Second, the surplus has to clear a hurdle before any use of it is worth doing, and the hurdle is not zero. At a deposit rateWhat a bank pays for money left with it for an agreed stretch of time. Rates change constantly. of 6.00 per cent before tax, which is this example's assumption and not a market figure, the surplus throws off Rs 4,80,00,000 across a year. Tax takes a quarter of that on the 25.0 per cent effective rate this company assumes for itself, leaving Rs 3,60,00,000. Rs 3,60,00,000 is the size of what any other use of the surplus has to beat.

What the Rs 60,00,00,000 stopped earning the moment it was spent

The company's own history supplies a spent amount to work on. At the end of the year two before the base year it bought back 75,00,000 of its own shares at Rs 80.00, spending Rs 60,00,00,000 and taking the count from 20,75,00,000 shares to 20,00,00,000.

The flattering reading of that is easy and it is arithmetically correct as far as it goes. Base year profit attributable to ownersThe slice of a group's profit that belongs to its own shareholders, after the share belonging to outside holders in a subsidiary has been taken out. came to Rs 1,38,00,00,000. Divide that by the 20,00,00,000 shares left once the purchase was done and each share carries Rs 6.90. Divide the identical profit instead by the 20,75,00,000 shares that would still be outstanding had nothing been bought, and each carries Rs 6.6506. 20.75 over 20.00 comes to exactly 1.0375, so exactly 3.75 per cent separates those two figures.

The honest reading puts back the cost nobody paid. The Rs 60,00,00,000 did not vanish when the shares were purchased, it was spent, and the return it would otherwise have earned is a cost of having spent it. On deposit at the assumed 6.00 per cent it would have produced Rs 3,60,00,000 before tax and Rs 2,70,00,000 after. Profit in that alternative world is Rs 1,40,70,00,000, spread over 20,75,00,000 shares gives Rs 6.7807, and against that figure the purchase lifted earnings per share by 1.76 per cent rather than 3.75.

The two bars repay careful reading. The lesson is not that somebody made a mistake with a calculator. Every step of the 3.75 per cent is right. The profit figure is right, both share counts are right, and the division is right. The missing line was never written down anywhere, so it was never available to be got wrong. The error has that shape every time: not a wrong number, but an absent one.

At what rate does the whole apparent gain disappear?

If putting back a forgone return at 6.00 per cent halves the measured gain, an obvious question follows: how high does the assumed return have to be before the gain is gone entirely? The rate is worth working out rather than guessing, and it comes out exactly round.

Earnings per share of Rs 6.90 on 20,75,00,000 shares needs profit of Rs 1,43,17,50,000. Profit of Rs 1,43,17,50,000 is Rs 5,17,50,000 more than the Rs 1,38,00,00,000 actually reported. Tax takes 25.0 per cent on this company's own assumption, so delivering that much extra after tax calls for Rs 6,90,00,000 before it. Against Rs 60,00,00,000 that works out at exactly 11.50 per cent. Assume anything above 11.50 per cent and the whole apparent gain from the purchase has gone. This company's weighted average charge for its capital, set out under the weighted average cost of capital, stands at 12.00 per cent. Nothing follows from the pairing on its own.

Where does a cost of capital sit in all this?

One observation ties the whole distinction to the cost of capital, and it takes one line. Sankalp Industrial Systems Limited has a weighted average cost of capital of 12.00 per cent. Nobody pays it. No account records it. No auditor confirms it. A cost of capital measures the return identical money would fetch somewhere else at identical risk, so it is an opportunity cost and nothing besides.

Everything odd about a cost of capital follows from that one fact. Two careful people are estimating what is available elsewhere rather than reading a bill, so they can build different ones for the same company. A cost of capital never appears in the accounts, even though it governs whether the company is worth anything. Profit is struck after everything the company paid, and the charge for capital was never one of those payments, so a company can report a profit every single year and still not have covered what its capital costs. How that gap is measured against profit is covered separately.

The two errors look nothing alike and push in opposite directions

Getting each side wrong distorts a decision in a different direction, so treating the two as one lesson loses half of it. Take them one at a time.

The first error is including the sunk cost, and it is the one everybody has a name for. Rs 52,00,00,000 has gone on fees, and the sentence that follows is always some version of: so much has already been spent that this now has to be made to work. The sentence changed nothing about the two futures being compared and changed the answer anyway. The argument keeps people in positions, projects and processes that no longer stand up on their own arithmetic. The tell is the tense: the argument is about the past while the decision is about the future.

The second error is excluding the opportunity cost, and it is quieter, more respectable and more expensive. The error sounds like this: the cash is already ours, so using it costs nothing. Rs 80,00,00,000 of surplus treated as free makes every use of it look better than it is. On the purchase worked above the difference between the two treatments is 3.75 per cent against 1.76 per cent, and the honest figure is less than half the flattering one. Push the assumed return to 11.50 per cent and the whole gain is gone.

Is there any conversation where a sunk cost is the right thing to talk about?

Yes, exactly one, and drawing the line around it prevents the opposite error of pretending the past never happened. A sunk cost belongs in the conversation that judges the decision which created it, and in no conversation about the decision now being taken.

Asking whether the Rs 52,00,00,000 of fees was worth agreeing to is a fair question about a choice that was made at a moment when the money had not yet moved. Asking is how a buyer learns anything at all, how a fee schedule gets negotiated harder next time, and how an adviser's value gets tested. Asking whether those same fees justify what to do next is a wholly separate question, and its answer never changes: no. Same rupees, two conversations, and only one of them is about a future that can still be chosen.

The distinction also keeps a decomposition honest. The minus 3.09 per cent line in the table above is doing the first job well. The same line would do damage the moment somebody quoted it in a meeting about whether to sell. Its words are about the past, and the room would be choosing between futures.