The Capital Budget: Dividing One Period of Capital

A capital budget is one period's capital stated as a single total and then divided: what the business generates, what cash it starts with, and what it can borrow inside a stated ceiling, set against every claim on that total. Claims that are not optional come out first. The remainder is the only money any competing use is actually arguing over.

The comparison sheet is already built. How a period's capital is counted, how each claim is sized in rupees and in periods, and how a return gets a stated basis are settled elsewhere. Turning that sheet into a budget is the narrower and more useful job. A budget is a document with one total at the top and a division underneath it. The sheet compares. The budget divides. Comparing and dividing are not the same job, and confusing them is how a company ends up with an excellent analysis and no decision.

So what is a capital budget, actually?

The word budget makes people picture a spreadsheet with two hundred rows and a tab for every cost centre. A capital budget is smaller and stricter than that. A capital budget is one period, one total, and a division of that total across the claims made on it. Everything else people put in a budget document is commentary.

The shape is identical at household scale, and the numbers there are small enough to hold in mind. The household version comes first. A household runs on one salary. Each month an amount arrives. An amount is already sitting in the savings account. And there is an amount the household could raise against the gold in the locker if it truly had to, up to a limit it has quietly set for itself. Nobody wants to pledge all of it. Salary, savings and pledgeable gold together are the month's total. Then the rent goes out, the school fee goes out, and the vehicle instalment goes out, and not one of those three was ever a proposal. The money left over is what the new scooter, the shop deposit and the cousin's wedding are competing for. The remainder, and not the salary, is what the argument at the dinner table is actually about.

A company works the same way at a different scale. Harivansh Packaging Limited, an invented packaging manufacturer, generates a flow during the year, holds a balance at the start of it, and can borrow up to a ceiling it has set on its own leverage. Some spending happens regardless. The money left after that spending is the contested money.

Notice what this makes a budget, and what it does not make it. Nobody in the room will dispute that Rs 342 crore, Rs 140 crore and Rs 354 crore add to Rs 836 crore. The room is going to argue about whether 2.0 times is the right ceiling, whether upkeep truly costs Rs 138 crore, whether a flow measured ahead of every movement in working capitalThe money tied up in stock, in what customers owe and in what suppliers are owed, all of which swings during a year. How it is measured is covered in the statements material; what matters here is only whether the figure exists. movement is honest. The sum is not where the disagreement lives.

How big is the total, and who gets to decide?

The total has three parts and one honest gap. Counting the parts belongs to the comparison sheet, covered separately. A budget lives or dies on whether the reader can see its total. The three-line sum is worth restating for that reason.

The first part is what the business throws off during the period. For Harivansh Packaging, earnings before interest, tax, depreciation and amortisation (EBITDA)Operating profit measured ahead of interest, ahead of tax and ahead of the write-down on assets that last many years. Constructed in the statements material and used here as given. of Rs 477 crore, less finance costThe interest and related charges paid to the people who lent the business money, taken out of profit before tax. Settled in the statements material. of Rs 60 crore leaves, and so does tax of Rs 75 crore, and what survives both is Rs 342 crore. Rs 342 crore is a flow. The money lands across the twelve months rather than on the first morning.

Everything after that figure leans on it. Precision about what it is matters. Rs 342 crore is not a published cash number. Nothing in this record reports cash movements at all, so the figure is assembled from three lines of the profit ladder, and it wears the word proxy every time it appears. Two things are still ahead of it: the swing in working capital, and the maintenance spending that has to come out. Rs 342 crore is the most generous honest reading of one year's output, never money sitting available.

The second part is simpler than anything else here: Rs 140 crore is sitting in the account on day one. A balance, not a flow. The balance needs no derivation, it is the most certain number in the whole budget, and it can be spent exactly once.

The third part is a decision wearing the costume of a measurement. Borrowing capacity is the distance between today's net debt and the level the company has publicly promised itself it will stay under. Net debt here stands at Rs 600 crore. Put the promise at 2.0 times EBITDA and Rs 954 crore is permitted, leaving capacity of Rs 354 crore. Shift the promise by half a turn in either direction, and this quantity swings by hundreds of crore. Nothing whatever changes inside the business. The budget therefore prints three versions of the capacity instead of one.

Now the gap. The absence matters more than its single blank cell suggests. Nothing published for this business reports how stock, customer balances and supplier balances moved during the year, nor what it spent on capacity. So a year that ties up an extra Rs 40 crore in those balances hands over materially less than a year that frees Rs 40 crore from them, and the record refuses to say which year this was. A budget records that absence as a line with no figure against it, rather than proceeding as though the missing number were nil. A believable swing dropped into the empty cell does three things at once: the total shifts, every fit judgement resting on the total shifts with it, and nothing in the document marks that a number was manufactured.

Which claims never had to be proposed?

Here is where a budget separates itself from a comparison sheet. The sheet lists every claim side by side so they can be looked at together. The budget takes some of them off the top before the looking starts.

Three kinds of spending land on the period without ever having been argued for. Keeping the existing plant capable of what it already does. Commitments signed in some earlier period that fall due inside this one. And repayments where a stated sum is owed on a stated date. No meeting has the power to reschedule one. Committed claims are not proposals, they do not compete, and a budget that puts them in the queue alongside the proposals has misdescribed the problem in its first line.

Go back to the household for a moment. The water pump gives out in April. Nobody pitched the pump, nobody built a case for the pump, nobody compared its return against the scooter, and the plumber gets paid regardless of what anybody thinks. Suppose the household instead writes down its total for the month and starts weighing the scooter against the shop deposit without first setting aside the rent and the pump. Such a conversation can be conducted with great care and still be about money that was never there.

For Harivansh Packaging the committed claim is maintenance spending, and here the record makes life awkward in a way worth being honest about. Nothing is published anywhere about what this business spends on its capacity, so the line has to borrow a figure from depreciation and amortisationAn accounting charge that writes down what was paid for durable equipment, a slice at a time, over the period it stays in service. Settled in the statements material. and stand at Rs 138 crore, with the word proxy written beside it in the cell itself. The two quantities are genuinely not the same thing. One looks backwards at equipment already bought and spreads its cost out; the other looks forward at what next year's upkeep will cost. The two figures tend to land near each other. The closeness is what makes the substitution tolerable and also what makes it easy to forget. Separating upkeep spending from growth spending is set out under maintenance and growth capital expenditure.

The repayment line is thinner still, and the budget admits it. Total borrowings of Rs 740 crore are known. No maturity split is published. How much of that pile matures inside these twelve months is therefore unknown. So the row goes on with an empty amount column and its reason written alongside, in the same shape as the working capital row above. Two blanks on one printed sheet read as carelessness until it becomes clear that a guess in either one would change which claims fit.

Now do the subtraction, one line at a time. Generation of Rs 342 crore less committed maintenance of Rs 138 crore leaves Rs 204 crore. Add the Rs 140 crore of cash held, and Rs 344 crore is what the competing uses are actually arguing over before anybody borrows a rupee. The business handles Rs 3,180 crore of revenue in the year. The contested money is roughly a ninth of that.

Why is an order not a ranking?

Once the committed claims are out, the remaining claims get put in an order. Ordering is the point at which most budget documents quietly stop being arithmetic and start being an argument, usually without anybody noticing that it happened.

Claims get ordered by two things, and neither of them is merit. The first is when the money is actually needed. A payment due in the second month sits ahead of one due in the eleventh, not because it is better but because a period runs in sequence and cash arrives across it. The second is which claims foreclose others. If committing to one use consumes the capacity a second use would have needed, then the second use has to be considered before the first is signed, or it will never be considered at all.

A queue is a scheduling device and a ranking is a judgement, and presenting the first as the second is how a budget begins making decisions it was never entitled to make. The two documents look identical. Both are a list of claims, one under another, with the biggest heading in the same font. A row's position means one thing on a schedule and another on a ranking, and the reader is almost never told which document they are holding.

The household version is sharp here. Suppose the school fee is due in June and the shop deposit is due in November. The fee goes first on the list. Nobody would say the list proves education matters more than the shop, and yet at a board table, with the same list drawn in a nicer typeface, somebody always does.

Which of the three constraints is actually binding?

A budget has a stopping point. Somewhere down the list there is a claim that cannot be funded, and something specific is stopping it. Naming that thing is one of the two or three genuinely useful outputs a budget produces. Everybody in the room already believes they know the answer, so the naming gets skipped constantly.

There are three candidates and they have to be tested one at a time.

Cash. Is there a rupee left in the account? For Harivansh Packaging there is. Rs 140 crore sits there untouched at the start of the period, and the generation keeps arriving. Cash is not what stops anything here.

The leverage ceiling. Would signing the next commitment push net debt past the line the company drew for itself and told people about? At 2.0 times EBITDA the permitted level is Rs 954 crore, net debt today is Rs 600 crore, and the capacity is therefore Rs 354 crore. A ceiling of the company's own choosing is not for that reason a soft one. A company that has told the market where it will stop and then steps past it has spent something that does not appear on any balance sheet.

A covenantA term written into a borrowing agreement that restricts what the borrower may do while the borrowing is outstanding. The terms of any real agreement are a matter for that document. in an existing facility. A borrowing agreement can contain a term that bites before either of the other two does, and when it does it is not negotiable at the pace of a budget meeting. The covenant used here is a term of an invented facility.

The binding constraint is very often not the one people quote: a company can be full of cash and out of headroom, or heavy with headroom and out of cash, and the budget's job is to say which. The two states look completely different from the inside and completely alike in a summary. A good budget therefore carries three tests and one named answer rather than a vague sentence about being constrained.

What happens to a claim bigger than one period?

Some claims are simply too large to be a decision about this year, and the rupee figure hides that completely. Rs 1,140 crore and Rs 140 crore are both numbers, both correct, and both look like the same kind of thing when set down side by side.

The cheapest fix available is to divide every claim by one period's generation and print the answer beside the rupee figure. Rs 342 crore is what this business produces in a year. So Rs 1,140 crore is 3.3 years of generation. Rs 540 crore of project equity is about 1.6 years. A repurchase or a repayment of Rs 140 crore is 0.41 years, and the committed maintenance is 0.40 years.

One caution attaches to that entire column, it leans the same way every time, and it is worth fixing in mind now. The denominator, Rs 342 crore, is the assembled proxy from earlier: nothing has yet been taken out of it for the swing in working capital, and nothing has been taken out for the Rs 138 crore of upkeep either. Every years figure in this guide is therefore a floor rather than a measure. Divide the same Rs 1,140 crore by the Rs 204 crore actually free after maintenance and it is 5.6 years, not 3.3, and the project equity moves from 1.6 years to 2.6. The truthful reading is a range with 3.3 years at its optimistic end, and this record does not carry what would be needed to close it. A floor is still useful: a claim that reaches into three periods at the most generous denominator available reaches into at least three.

Read that column and the character of each claim changes. A claim larger than one period's generation is a claim on periods that have not been budgeted yet, made by people who will not be in the room when the last of it is paid for. That is not an argument against large claims. Businesses have to make them. The years column is an argument for saying out loud that a three-year claim is being approved inside a one-year document. Saying it out loud changes how carefully the document gets read.

Expressing large claims in years of generation is close to free, takes one extra column, and almost nobody does it. The reason is uncomfortable: the column makes the biggest proposal look the most demanding, and the biggest proposal usually has the most senior sponsor.

The budget for Harivansh Packaging Limited, written out

Here is the whole thing for one year, standalone, with every figure labelled and the two absences left as absences.

Four things this record does not carry, listed once so that no row below has to explain itself twice: there is no cash flow statement, no capital expenditure figure, no dividend or payout of any kind, and no run of prior years to read a trend from. Each of those shapes a cell in the tables that follow, and in every case the budget prints the absence rather than a number standing in for it.

Put the borrowing capacity back on top of that and the whole total, after the one committed claim anybody could size has left it, is Rs 698 crore: Rs 836 crore of three-part total less the Rs 138 crore of upkeep. Rs 344 crore of it needs nobody's permission, and the remaining Rs 354 crore arrives only by moving net debt.

Now set the remaining claims against that. Each sorts into exactly one of three boxes, and the sort is the budget's real work.

Two notes on how those rows were sized. Sizing is where quiet inventions get in. The two Rs 140 crore rows are pinned to the cash balance for one reason and no other: not a rupee of distribution appears anywhere in this record, in any form or from any year, so no policy figure exists to size them with and the balance on hand is the only anchor that does not require inventing something. Rs 1,320 crore is enterprise value. The acquisition row runs on Rs 1,140 crore instead, and equity valueWhat actually reaches the sellers once the debt inside the business being bought is taken off the enterprise value. The bridge between the two was settled in the transaction valuation material. of Rs 1,140 crore is what the acquirer hands over. Quoting the larger figure as the price paid is the commonest error made about transaction numbers anywhere.

The finding, and it is not about the target

Look at the last row again. The acquisition is not a capital budget item at all. Nothing in this document can accommodate it, and no amount of reordering the claims above it will change that. The acquisition is a proposal to move the leverage ceiling, and calling it that is more useful than any further debate about where it belongs on the list.

Quote the move on whatever base the ceiling itself was struck on, or the sentence carries no meaning. The 2.0 times ceiling is a limit written against Harivansh Packaging alone: its own net debt, its own EBITDA, nobody else's. Hold that base fixed and follow the funding through. Borrowings of Rs 740 crore take on Rs 1,000 crore more, the Rs 140 crore of cash is spent down to nothing, and net debt for the acquirer by itself arrives at Rs 1,740 crore. Set against its own Rs 477 crore, that reads 3.65 times. The year opened at 1.26 times. The request, then, is permission to run at 3.65 times against a ceiling of 2.0 times.

A second reading is equally honest and has to wear a label wherever it shows up. A purchase of all the shares in Sundarban Polymers Private Limited brings that seller's own Rs 180 crore of net debt onto the combined books, taking the consolidated figure to Rs 1,920 crore against combined EBITDA of Rs 609 crore. The combined figure reads 3.15 times. Consolidated leverage is a different measure of a different thing, and it is not the measure the 2.0 times ceiling was ever struck on. Setting a consolidated ratio against a standalone ceiling is comparing two quantities that were never the same quantity.

The binding constraint has to be stated in words. The constraint is not cash: the Rs 140 crore is untouched and would still be untouched after the two Rs 140 crore claims were considered. Appetite is not the constraint either, and neither is the quality of the target. The target's quality is a separate question. The constraint is the ceiling. And whether to move a ceiling is a different question from whether an asset is worth buying, asked of different people, on a different timetable, with a different set of consequences if it goes wrong.

What does a budget produce, and what can it never produce?

A budget's output is short and worth listing exactly. A total, built rather than asserted, with all three parts visible. The committed claims removed before allocation. Every remaining claim sized in rupees and in years. Each claim sorted into fits, fits only by borrowing, or does not fit. The binding constraint named. And every absence left visible as an absence.

The choice is the one thing a budget cannot produce. The claims differ in kind, in certainty and in what they foreclose, and no arithmetic in a budget resolves any of those three. They differ in kind because returning capital to holders and buying a business are not the same species of act. The claims differ in certainty because Rs 140 crore repaid is a known outcome and Rs 1,140 crore of acquisition is a range with a long left tail. And they differ in what they foreclose. The repayment restores capacity; the acquisition consumes several periods of it.

So a budget presented as a decision has smuggled a judgement into a document that does not contain one. The tell is usually formatting: a summary line at the bottom, a row highlighted in a different colour, a claim placed first with no note saying why. None of those is arithmetic, and all of them are read as though they were.

What reopens a budget in the middle of a period?

A budget is written once and then reality arrives. Three things reopen it, and only three.

The first is a claim arriving that exceeds the total. Not a claim that is inconvenient or badly timed, but one the document has no room for at any ordering. The acquisition is exactly that, and recognising it as a reopening rather than as a line item is more useful than a long argument about where it fits. A budget can absorb a claim it did not expect. A budget cannot absorb a claim larger than itself, and pretending otherwise produces a document that has stopped describing anything.

The second is a constraint moving. If the leverage ceiling is reset from 2.0 times to 2.5 times, capacity goes from Rs 354 crore to Rs 592.50 crore and the total from Rs 836 crore to Rs 1,074.50 crore, and every fit judgement below has to be run again. Nothing about the business changed. A sentence in a policy changed, and the arithmetic beneath it moved by Rs 238.50 crore. A moving ceiling cuts the other way too: reset down to 1.5 times, it leaves Rs 115.50 crore of capacity, and the project equity that fitted by borrowing no longer fits at all.

The third is a committed claim turning out to be larger than it was booked at. Consuming contested money without any meeting taking place, an underbooked commitment is the quietest of the three and the most damaging. Maintenance carried at Rs 138 crore that actually costs Rs 180 crore has taken Rs 42 crore out of the Rs 204 crore of free generation, a fifth of it, and nobody proposed the increase or approved it. The same happens when a commitment agreed in an earlier period lands heavier than booked, or when an earn-outA further amount payable to sellers only if the business bought reaches an agreed level of performance after completion. How one is drafted and measured belongs to the documentation material. agreed in a past transaction becomes payable.

Reopening a budget is not a failure of the budget; refusing to reopen one is. A document written against a total that no longer exists keeps producing fit judgements that are all wrong in the same direction, and it looks exactly as authoritative as it did on the day it was correct.