DuPont — what actually drives return on equity

The problem it solves

Return on equity is one number carrying at least three different stories, and on its own it cannot tell you which one you are reading.

A 25% ROE might mean the company earns a fat margin on every sale. It might mean thin margins turned over very fast — the supermarket model. Or it might mean modest returns amplified by a great deal of borrowed money.

Those are not the same business, and they do not deserve the same multiple.

The identity

The three-step form multiplies out to exactly ROE, by construction:

ROE = net margin × asset turnover × equity multiplier

• Net margin (net income ÷ revenue) — profitability per sale
• Asset turnover (revenue ÷ assets) — how hard the asset base works
• Equity multiplier (assets ÷ equity) — how much of those assets is funded by someone other than shareholders

Because it is an identity rather than a model, it cannot be wrong. It can only be uninformative — which is why the five-step form exists.

The five-step form

The extended version splits net margin further, separating out the two non-operating claims on profit:

ROE = tax burden × interest burden × operating margin × asset turnover × equity multiplier

Now leverage appears twice, and in opposite directions. The equity multiplier raises ROE; the interest burden (pre-tax profit ÷ operating profit) lowers it, because borrowing costs money. That pairing is the whole value of the five-step version: it shows whether the debt is paying for itself.

A company whose equity multiplier is climbing while its interest burden falls is running its ROE on borrowed money, and the headline number can rise throughout. What the decomposition cannot tell you is whether that was a good trade: if the borrowed capital earns more than it costs, a rising multiplier and a falling interest burden are exactly what a profitable, accretive financing looks like. The split shows you where the return came from, and you still have to compare the return on that capital against its cost.

Reading it over time

One year of DuPont tells you the shape of the business. Several years tell you what is changing, which is the more useful question.

Margin improving with turnover flat is a pricing or cost story. Turnover improving with margin flat is an efficiency story. Both flat with ROE rising is a leverage story, and it is the one that reverses in a bad year.

Pythia computes both forms from the filed lines rather than from stored ratio columns, because rounding a ratio and then multiplying breaks the identity by a few basis points — and an identity that does not close exactly stops being evidence and becomes an estimate.

The limits

The decomposition is arithmetic, so it explains nothing on its own. It tells you which lever moved, never why, and never whether the movement is durable.

It also inherits every distortion in its inputs. Equity that has been reduced by years of buybacks makes the multiplier large and the ROE flattering; in the extreme, negative equity makes ROE meaningless rather than infinite. When equity is negative or earnings are negative, the honest output is a refusal, not a number.