ROE and ROCE — return on whose money

The only difference is the denominator

Both ratios ask the same question — how much profit does this business make per rupee tied up in it — and disagree about whose rupees count.

Take the profit left over after the company has paid interest to its lenders and tax to the government, and divide it by what the shareholders own on the books: share capital plus every rupee of profit retained over the years. That is return on equity, ROE.

Now take the profit the business earns before it pays either of them, and divide it by all the long-term money the business is using — shareholders' funds plus borrowings. That is return on capital employed, ROCE.

One denominator excludes debt. The other includes it. Everything that follows in this article — why a company can improve one ratio by taking an action that makes it riskier, why the two disagree, why neither travels well between businesses — falls out of that single difference in what sits below the line.

Which is also the reason to read them as a pair. Alone, ROE tells you what shareholders earned without telling you how much borrowed money was standing behind them. Alone, ROCE tells you what the assets earned without telling you how much of it the lenders took.

Each numerator belongs to its own denominator

The two profit figures are not interchangeable, and swapping them is the most common arithmetic error in this pair.

Interest is the price of debt. If debt is out of the denominator, as in ROE, then the lenders' share must be out of the numerator too — which is why ROE uses profit after interest. If debt is in the denominator, as in ROCE, the interest has to be added back, because the money that paid it is sitting in the denominator and the return it produced has to be counted somewhere.

Divide profit after tax by capital employed and you get a number that credits the business with none of what its borrowed money earned while charging it for all of it. Wherever the company carries debt it sits below both of the real ratios, and it sinks further the more the company borrows — none of which has anything to do with the business.

One further mismatch is built into convention and worth holding on to. ROE is measured after tax; ROCE is usually measured before it. They are not on the same scale, so the gap between the two printed numbers is not the benefit of leverage, or of anything else. Any comparison between them has to put them on one basis first — a point this article returns to once the arithmetic is on the table.

How borrowing lifts ROE and leaves ROCE alone

The cleanest way to see it is to hold a business completely still and change only how it is funded.

Here is an illustrative company. Every figure in it is invented to make the arithmetic visible, and none of it describes a real business or a typical one. It uses capital of ₹400 crore, earns ₹80 crore a year before interest and tax, borrows at 10%, and pays tax at 25%. The assets, the customers and the operating profit are identical in all three rows — only the funding mix changes.

ROCE does not move, because neither half of it moved: the same ₹80 crore of operating profit over the same ₹400 crore of capital, however that capital was raised. ROE two and a half times itself, from 15% to 37.5%, and the business did not get better at anything. Profit after tax actually fell in rupee terms, from ₹60 crore to ₹37.5 crore. It is being divided among far fewer shareholder rupees.

The relationship compresses into one line, and it is worth carrying because it explains every row above:

ROE = [ROCE + (ROCE − borrowing rate) × debt-to-equity] × (1 − tax rate)

Check the middle row against it. ROCE of 20, less a borrowing rate of 10, is 10 percentage points of spread; debt equals equity so the multiplier is 1; that gives 30, and three-quarters of 30 is 22.5%. The bottom row has three rupees of debt per rupee of equity, so the same 10-point spread is counted three times: 20 plus 30 is 50, and three-quarters of 50 is 37.5%. The multiplier in the middle of that formula is the debt-to-equity ratio, which is where its own construction and its own traps are set out.

Read the formula slowly and it says something the ratio alone never does. Borrowing does not add return. It adds a multiple of a spread. The spread is the only thing the business earned; the multiple is a decision about how much of somebody else's money to stand on. Whether standing on it is prudent is a separate question, taken up in good debt versus bad debt.

The point where borrowing stops helping

Borrowing stops helping at the point where ROCE equals the borrowing rate. Above it, every borrowed rupee earns more than it costs and the surplus belongs to the shareholders; below it, the shortfall is theirs too. The pivot is exact rather than approximate, and the table shows why.

Run the same three funding structures through three different operating years. The first column is the year already worked through, at ₹80 crore of operating profit. The second drops it to ₹40 crore and the third to ₹32 crore. Nothing about the funding changes; only the business's own return on its capital does.

Look at the middle column. Three companies with wildly different balance sheets return exactly the same 7.5% to their shareholders, and the reason is not a coincidence of the numbers chosen. ROCE is 10% and the borrowing rate is 10%, so every borrowed rupee earns precisely what it costs and contributes nothing either way. Put ROCE equal to the borrowing rate in the formula above, the spread term goes to zero, and the debt-to-equity multiplier has nothing left to multiply.

That is the pivot, and it is the single most useful thing the pair will tell you. On one side of it, leverage multiplies a real operating return into a larger shareholder return. On the other, it multiplies a shortfall. In the right-hand column the three-quarters-debt company earns 8% on its capital, pays 10% on most of it, and hands its shareholders 1.5% — a quarter of what the unlevered version of the same business managed.

The trade-off is not symmetric in practice, and this is where the arithmetic stops and the real world starts. Operating profit is uncertain; interest is contractual. The 8% column is also the column in which a lender's covenant is tested and the debt has to be refinanced, and the rate it is refinanced at need not be the old one — a floating rate resets on a schedule the borrower does not control, as interest rates explained sets out. Leverage widens the distribution of outcomes and shortens the time available in the bad tail.

There is a usable test hiding in the identity. Compare ROE with ROCE reduced by the tax rate — at a 25% rate, three-quarters of ROCE. If ROE sits above that figure, the company's capital is earning more than its debt costs. If it sits below, the borrowing is costing more than the assets are producing, and the gap is coming out of the shareholders.

Two conditions on that test, and both are easy to forget. A company with no debt lands exactly on the line whatever its operating return — the top row of the first table earns 20% on its capital and still prints ROE of precisely three-quarters of it — so on an unlevered company the test is silent, not negative. And it sets two computed numbers against each other: if they came from different sources, or the tax rate used is not the rate the company actually paid, the gap is measuring the two constructions as much as the business underneath them.

Three different reasons ROE can rise

A rising ROE is treated almost everywhere as evidence of a better business. It is evidence of one of three things, and only two of them are about the business.

Multiply three fractions together: profit divided by sales, sales divided by assets, and assets divided by equity. Sales cancels against sales and assets against assets, so what is left is profit divided by equity — ROE, arrived at the long way round. Splitting it like that is the DuPont decomposition, and it exists because the three fractions describe three unrelated things.

• Profit margin — how much of each rupee of sales survives to the bottom line. A pricing and cost story.
• Asset turnover — how many rupees of sales each rupee of assets generates. An efficiency story: the same factory running two shifts instead of one.
• The equity multiplier — assets divided by equity, which is simply how much of the asset base was funded by somebody other than shareholders. A financing story, and not a story about the business at all.

Put the illustrative company through it. Say it makes sales of ₹600 crore on its ₹400 crore of assets. In the all-equity version, profit after tax of ₹60 crore is a 10% margin, turnover is 1.5 times, and the multiplier is 1 because equity funds everything: 10% × 1.5 × 1.0 gives 15%.

In the half-debt version, profit after tax of ₹45 crore on the same sales is a 7.5% margin, turnover is unchanged at 1.5 times, and the multiplier has doubled to 2. That is 22.5%. The margin got worse and the ROE got better, because interest ate into the first term while the third term doubled.

Here is the mistake to be able to recognise, stated precisely enough to catch yourself making it: reading a multi-year rise in ROE as operational improvement without looking at which of the three terms moved. Margin flat, turnover flat, multiplier climbing is not a business getting better — it is a balance sheet getting heavier, and the ratio reports the two identically.

ROCE is harder to flatter this way, because the equity multiplier has no equivalent in it. Its denominator already contains the borrowed money, so adding more borrowing adds to the bottom of the fraction at the same time as it adds to the assets on top. Harder is not impossible — the rest of this article is about the ways both denominators can be moved without a business changing at all.

“Capital employed” is not one number

Before comparing any two published ROCE figures, there is a prior question that almost nobody asks: computed how.

Shareholders' funds is a line in the accounts, so the ROE denominator is at least well defined. Capital employed is not a line anywhere. It is a construction, and at least three constructions are in common use:

• Total assets less current liabilities.
• Shareholders' funds plus total borrowings.
• Net fixed assets plus working capital.

Each of those can be taken at the year-end figure or averaged across the year, with or without cash netted off, with or without short-term borrowings, with or without intangibles. The permutations run into dozens, and they do not produce approximately the same answer — on a company holding a large cash pile or carrying heavy seasonal working capital, two defensible constructions can differ by a third.

The consequence is practical. Two websites showing different ROCE for the same company are usually both right and computing different things, and a screener's ranking is a ranking on whichever construction that screener chose. A ROCE figure whose formula you cannot see is a number of unknown make.

Cash deserves its own line, because it moves the two halves in opposite directions. Money sitting in deposits is part of capital employed under most constructions, but the income it earns is treasury income, which most constructions of operating profit exclude. So a cash-rich company is charged for the cash in the denominator and given no credit for it in the numerator, and its ROCE reads low for a reason that is an accounting convention rather than a fact about its operations.

What can move the denominators without the business moving

Every item here changes a ratio while leaving the operations untouched. That is what makes them worth listing: none of them shows up as a change in the business, and all of them show up as a change in the number.

• Buybacks and large dividends — both hand cash back and shrink shareholders' funds. The numerator is unchanged and the denominator is smaller, so ROE rises mechanically. Capital employed falls too, so ROCE also rises, though by less where the cash was funded partly by debt.
• Write-offs and impairments — a goodwill impairment or a one-time asset write-down cuts reserves in the year it happens, and the smaller equity base stays smaller. Every subsequent year's ROE is computed against it. Call this the vanishing denominator: the ratio improves permanently because value was destroyed once.
• Accumulated losses — the same effect running for years. A long loss-making history erodes reserves until a modest profit lands on a tiny equity base and prints an impressive ROE. Past the point where equity turns negative the ratio stops meaning anything at all, and a loss divided by a negative denominator will helpfully report as a positive percentage.
• The age of the assets — a plant bought two decades ago and depreciated down to near nothing carries a small book value, so the same output divides by a much smaller denominator than an identical plant built this year at today's prices. The older business looks more productive purely because inflation has moved the replacement cost away from the recorded one.
• Revaluation — where assets are carried at revalued amounts rather than at cost, the reserve created inflates both denominators and depresses both ratios, again with no operational change.
• Acquired goodwill — a company that bought its growth carries the premium it paid inside capital employed. The company that built the same capacity itself carries only the cost of building it. Comparing their ROCE compares two different accounting histories more than two operating records.
• Year-end against average — a company that raised fresh equity a fortnight before the year-end divides a full year of profit by a denominator that includes money it has barely had. The following year, the same business reports a much worse ratio.
• Where a lease sits — a long-term lease can be treated as a rent expense outside the balance sheet, or recognised as an asset with a matching liability on it. Under the second treatment capital employed is larger, and the rent is split into depreciation and a finance cost, which lifts operating profit. Both halves of ROCE move, in opposite directions, for a lease-heavy business that has changed nothing.

None of these is an accounting failure. Each is a defensible policy doing its job. The point is narrower and harder to dismiss: a ratio inherits every judgement made in constructing its denominator, and those judgements were not made with your comparison in mind.

And what can move the numerators

The top halves are no more solid, and they fail in a different way — not by slow drift, but in single years that then anchor a comparison.

• One-off items — a gain on selling land, an insurance settlement, a litigation provision reversed. Profit after tax carries all of it, so a single year's ROE can describe an event rather than a business. The tell is a ratio that jumps and returns.
• The tax position — ROE is measured after tax, so a company setting off carried-forward losses, or reversing a deferred tax provision, shows a higher ROE than an identical company paying full tax. That is a fact about its tax history, not its operations, and it reverses when the shelter runs out.
• What counts as operating — dividends received, treasury gains and the share of profits from associates sit in a grey zone. Include them in the ROCE numerator and the ratio rises; exclude them and it falls. Both choices are made, by different sources, for the same company.
• Depreciation policy — lengthening an assumed useful life reduces the annual depreciation charge, which raises operating profit, while also leaving a higher net book value in the denominator. The two effects push ROCE in opposite directions and neither is visible in the ratio.
• Capitalising rather than expensing — development spending routed through the balance sheet instead of the profit and loss account raises profit this year and raises capital employed at the same time. A company that expenses the same spending reports lower profit on a lower base.
• Consolidated against standalone — consolidated profit includes subsidiaries the parent does not wholly own, and whether the minority's share is stripped out of the numerator, the denominator, both or neither changes the answer.

A working habit falls out of the two lists. Read a single year's ROE or ROCE as a measurement with error bars rather than a fact, and read a run of five or six years for the shape. A step change in either ratio is a question, not a finding, and the answer is usually somewhere in the notes to the accounts rather than in the ratio itself.

Return on the company's money, not on yours

One confusion is worth killing outright, because the words invite it. Return on equity is not the return you get.

The denominator is book equity — what the shareholders' funds are recorded at in the accounts. What you pay is the market price, and the two are related only by whatever the market currently thinks. Divide profit by the price you paid rather than by the book value and you get the earnings on your own outlay, which is ROE divided by the price-to-book multiple — the multiple itself being the subject of PE and PB ratios.

Illustratively: a business earning 20% on its book equity, bought at four times that book value, is producing 5% on the money actually handed over. The company is unchanged in both readings. The arithmetic is not a valuation method and it is certainly not a threshold — it is one line showing that a ratio measured on the balance sheet says nothing about what a share costs.

The same idea underneath net worth as a personal measure — that what you own is recorded at a value someone assigned, not at a value someone paid — runs through net worth explained.

Where the pair stops working, and why no threshold appears here

Three boundaries, and then the sentence this article has been building towards.

The first is a whole sector, and there the pair does not bend but breaks. For a bank or an NBFC, borrowed money is not how the assets were funded — it is the raw material of the product. Interest paid is a cost of what is being sold, in the way that steel is a cost to a carmaker. Adding it back to construct an operating profit, and putting deposits and borrowings into capital employed, produces a ratio measuring nothing recognisable. These businesses are read on other measures, and their leverage is governed by capital rules rather than by choice.

The second boundary is capital intensity. A services business that needs almost no assets can show a very high ROCE on a small base, and a business building assets that will take years to produce revenue shows a low one throughout construction. The ratios are describing how much capital each model requires, which is a fact about the industry before it is a fact about the company.

The third is timing. Capital goes in before profit comes out. A company midway through an expansion carries the full denominator and only part of the numerator, and its ratios fall for the duration — which is indistinguishable, in the ratio alone, from a business getting worse.

Which is why no level is called good anywhere in this article, and the omission is deliberate. The cost of capital differs between companies, the capital a business model requires differs between industries, and the accounting policies underneath both numbers differ between filings. A single cut-off applied across all three is arithmetic dressed as judgement, and it would be a recommendation wearing a number. FNOTrader is not a registered investment adviser or research analyst, and nothing here is advice about any security.

What the pair does support is a sequence of questions. Is the return coming from margin, from how hard the assets work, or from how much was borrowed. Is the operating return above the borrowing rate or below it. Did the denominator change because the business changed, or because something was written off, bought back or revalued. Those questions survive a change in accounting policy, and a threshold does not.

Checking any of this against real filings

Every distortion above is visible in the accounts of the company it applies to, and invisible in any ratio computed from them. So the work is reading two things together: the number, and the note that explains what went into it.

FNOTrader's Stocks app screens and ranks the listed Indian universe on stated fields, and its backtesting layer runs factor rankings across roughly 2,390 stocks and 17 NSE sector and size indices, reporting the outcome in rupee terms rather than as a score. How the screener is built and what each field means is set out in the stock screener guide.

One caution that follows directly from this article, and it applies to every screener including ours: a ranking is only as good as the formula behind the field it ranks on, and the formula is a choice somebody made. Read the field definition before ranking on it, and expect a ratio from one source to disagree with the same-named ratio from another.

Common questions

Continue reading

More in Company Analysis · App: Stocks · Definitions: glossary · Free tools: calculators · All: every article