ROE limitations

Judgment starts where these ratios stop. Each one built so far is correct arithmetic on reported numbers, and each can mislead if you forget what it's blind to. Four structural limits follow, each with its fix.

How does leverage inflate ROE?

Limit 1 is that leverage inflates ROE, and DuPont alone can't tell you whether the leverage is safe. Write the mechanism down. Let R = after-tax operating return on assets (NOPAT ÷ average assets, NOPAT = EBIT × (1 − t)) and i = after-tax cost of debt. Profit to shareholders is what assets earn minus what debt costs, NI = R × (D + E) − i × D. Divide by E:

ROE = R + (R − i) × D/E

Every rupee of debt adds the spread (R − i) to shareholders' return, as long as the spread is positive. Put numbers on it with synthetic "Deccan Forge" (₹ cr). Assets are 1,000 and EBIT is 150, with tax at 25%, so NOPAT is 112.5 and R = 11.25%. Debt costs 9% pre-tax, so i = 6.75%:

The same 2.0× D/E that manufactures a 20% ROE in the good year erases shareholders entirely in a downturn a debt-free twin shrugs off. The spread (R − i) flipped negative, and the multiplier works both ways. Check the D/E = 2 good-year row longhand. D = 667 and E = 333. Interest = 9% × 667 = 60, EBT = 150 − 60 = 90, tax is 22.5, NI = 67.5, and ROE = 67.5 ÷ 333 = 20.25%. So leverage adds return only when R exceeds i, and it always adds risk. DuPont displays the leverage but can't judge it. Judging it is the job of the coverage and survival tests in Ratio Toolkit II.

Why can margins and book capital mislead?

Limit 2 is that margins mean nothing without capital intensity. A 25% operating margin sounds superb. But if producing it takes assets of four times revenue (turnover 0.25×), the pre-tax return on assets is 25% × 0.25 = 6.25%. That's likely below the cost of the capital employed, so the "profitable" company destroys value as it grows. Now flip it round. Walmart's 4.3% operating margin on 2.66× turnover yields 11.4% pre-tax on assets, which makes it a better business than the margin star. The product margin × turnover (= ROA, at the operating level) is always where wisdom begins. Either factor alone is marketing. Even ROA uses total assets, though, and that raises a deeper problem. Total assets and book equity are accidents of history. Two companies you've already met show the distortion:

• Infosys holds roughly a quarter of its balance sheet in cash and liquid investments (illustrative, ~FY24/FY25 scale). That treasury earns 6 to 7%, which drags measured ROA and ROE below the return of the operating engine. The business is better than the ratio.
• HUL's two ROEs. In FY19, HUL earned roughly ₹6,000 cr on about ₹7,700 cr of equity, an ROE around 80%. Today it earns ~₹10,300 cr on ~₹51,000 cr, an ROE around 20%. Did the business collapse? No. The 2020 merger with GSK Consumer Healthcare put roughly ₹45,000 cr of goodwill and acquired brands (Horlicks, Boost) onto the balance sheet, and the shares issued to pay for them went into equity at market value. Dove and Surf Excel, organic brands built through decades of expensed advertising, are on the books at approximately nothing. Horlicks, which was bought, is carried at fair value. Same company, same margins. One acquisition, and ROE is divided by four. (Illustrative, modeled on the actual FY19 vs FY24 filings. Pull both and see it for yourself.)

The fix is to measure the capital actually invested in operations. Strip out surplus cash, decide deliberately how to treat goodwill, and separate operating from financing. That's ROIC, and it gets a module of its own. Treat today's ROA and ROE as the field instrument and ROIC as the laboratory measurement.

What else can the ratio hide?

Limit 3 is that the numerator is accounting profit. It isn't cash, and it isn't necessarily repeatable. One-off gains, "other income," a tax credit or an under-depreciation policy all flow straight into net margin and ROE. The 5-step catches the tax and interest varieties. A CFO/NI test catches profits unbacked by cash, benchmarking strips one-offs before comparison, and the forensic weeks handle the deliberately dishonest cases. Until then, ask one question of every striking ratio. What's actually in the numerator?

Limit 4 is that a ratio is a photograph and a business is a film. A cyclical at its peak shows glorious margins and ROE just when the next five years will be worst. That's the commodities trap, and the commodities module deals with it. A harvesting company that has cut advertising and R&D shows expanding margins while its future walks out of the building. So always look at five to ten years, and always against peers. Benchmarking turns that into a routine.