A Simple Formula With a Fragile Denominator
ROE = net income ÷ shareholders’ equity
The Federal Reserve uses average equity capital in the denominator rather than a period-end snapshot, which smooths the distortion when equity moves sharply during the year. The numerator is uncontroversial. The denominator is where the trouble lives.
Book equity is not the value of the business. It is a residue of bookkeeping entries, and Regulation S-X sets out what those entries are: additional paid-in capital, other additional capital, retained earnings, and the contra-account for treasury stock. Nothing in that list is a market valuation.
Two companies with identical operations and different financing histories will report different equity, and therefore different ROE.
How Boeing’s Equity Almost Vanished
Buybacks accumulate against retained earnings
When a company repurchases its own shares, the cost is recorded as treasury stock, a deduction from equity. Do it for long enough and the subtraction approaches the accumulated profits it is being netted against.
Boeing repurchased $9.0 billion of stock in 2018 and $9.2 billion in 2017, and at December 31, 2018 still had $20 billion authorised under a plan approved that month. By that date treasury stock stood at $52,348 million against retained earnings of $55,941 million, leaving total shareholders’ equity of $339 million, down from $1,656 million a year earlier.
What the next year showed
One year later Boeing reported total shareholders’ equity of negative $8,617 million and a net loss of $636 million, having suspended repurchases under that $20 billion plan. The spectacular ROE and the undefined one describe the same company fourteen months apart.
Return on assets tells the more stable story. Boeing’s $10,460 million of 2018 earnings sat on $117,359 million of total assets, an ROA near 8.9%. That figure is unremarkable, which is the point: the assets did not disappear when the equity did.
Whenever ROE looks extraordinary, check book equity against total assets before crediting management. A denominator that has been hollowed out by accumulated buybacks produces a large ratio for reasons unconnected to operating performance, and return on assets will show it immediately.
When Equity Goes Negative, the Ratio Stops Existing
McDonald’s closed 2025 with a shareholders’ equity deficit of $(1,791) million. Cumulative treasury stock of $79,316 million exceeded retained earnings of $70,282 million, and the company nonetheless earned $8,563 million of net income that year.
Dividing $8,563 million by negative equity produces a negative ROE for a highly profitable business. The ratio has not detected a problem; it has ceased to function. McDonald’s reports after-tax return on invested capital instead, and discloses total debt at 105% of total capitalization at the end of 2025, against 111% in 2024 and 114% in 2023 — figures above 100% only because the equity component is negative.
The distinction between the two cases is worth holding onto. Boeing's equity collapsed alongside a genuine operating crisis, so the undefined ratio in 2019 coincided with real trouble. McDonald's negative equity reflects decades of returning capital from a business that never stopped performing.
The ratio breaks identically in both cases, and its breaking therefore tells you nothing about which situation you are in.
This is not a distressed company. It is a mature, cash-generative business that has returned more to shareholders over its life than it has retained. Book equity simply stopped being a meaningful denominator somewhere along the way.
Where the Return Actually Comes From
ROE can be decomposed into three drivers: how much profit each sale produces, how efficiently assets generate sales, and how much the balance sheet is levered. Two companies can report the same ROE with entirely different mixes.
Written out, the identity is straightforward: net margin times asset turnover times the ratio of assets to equity. The first term is a pricing and cost question, the second an operational one, and the third is purely financial.
The practical use is diagnostic. A retailer and a luxury goods house can post the same ROE from opposite directions: thin margins on rapid turnover in one case, fat margins on slow turnover in the other. Neither is better, but they are different businesses, and the single ratio conceals which one you are looking at.
The third driver deserves the most skepticism, because leverage raises ROE without improving anything. Borrowing to buy assets increases the numerator’s earning base while leaving equity unchanged, so the ratio rises. It also increases the loss in a bad year, symmetrically and invisibly, until the bad year arrives.
The scale of the buyback effect
Repurchases are not a marginal influence on this ratio. The SEC put the dollar volume of share repurchase activity at nearly $950 billion in 2021, conducted largely within the safe harbour of Rule 10b-18, which shields issuers from anti-manipulation liability when they buy their own stock within specified manner, timing, price and volume conditions. Across the market, ROE denominators are shrinking continuously.
Banks Show the Gap Most Clearly
Financial institutions make the leverage effect impossible to miss, because they run the highest leverage of any sector and disclose both ratios side by side.
JPMorgan Chase’s 2025 annual report shows return on common equity of 17%, return on tangible common equity of 20%, and return on assets of 1.29%. The same profits, expressed against two denominators, differ by roughly thirteen times. The bank closed the year with $4,424,900 million of assets against $362,438 million of total equity, about 12.2 times leverage.
The pattern holds industry-wide. In the first half of 2025 the US banking system produced a return on average assets of about 1% against a return on equity of about 10.5%, per the Federal Reserve’s supervision report, which also puts median ROE for large US banks at 13% in the third quarter of 2025, up from 12% in the second.
Why the Denominator Is a History, Not a Value
It is worth being concrete about what book equity actually records, because the misunderstanding drives most misreadings of this ratio.
Equity is the accumulation of everything that has ever happened to the company's capital: money raised from investors, profits kept rather than distributed, losses absorbed, and shares bought back. It is a running total of transactions, kept at the values recorded when each occurred. Assets bought decades ago sit at historical cost less depreciation, not at what they would fetch today.
That is why a company with valuable brands, established distribution or a workforce it spent years assembling can carry modest book equity: none of those appear on the balance sheet unless they were purchased from someone else. And it is why a company that has returned more cash over its life than it retained can carry negative equity while operating normally.
The ratio therefore divides a current-period flow by a cumulative historical stock. When those two are roughly proportionate, the answer is informative. When the denominator has been shaped by decades of financing decisions unrelated to this year's performance, it is not.
Tangible Equity and the Goodwill Question
JPMorgan’s 17% and 20% differ because the second excludes goodwill and intangibles. Book value per share was $126.99 at the end of 2025 against tangible book value per share of $107.56, so roughly $19 per share of the equity base consists of intangibles rather than hard assets.
Return on tangible common equity is the higher figure by construction, since removing intangibles shrinks the denominator. It is the more meaningful measure for an acquisitive company, where goodwill from past deals can inflate book equity and depress ROE without anything about current operations changing. When a company promotes one and not the other, the reason is usually which number is larger.
A short checklist before trusting any ROE. Is book equity a sensible fraction of total assets, or has it been hollowed out? Is the company acquisitive, so goodwill is inflating the denominator? And what is the leverage ratio, since ROE rises with borrowing whether or not the business improved?
The Conditions That Break It
- Sustained buybacks destroy the denominator. Boeing at 3,000% and McDonald’s at negative equity are the same mechanism at different stages.
- Leverage inflates it. Borrowing raises ROE without improving operations, and raises losses in the same proportion.
- Sector-to-sector comparison is meaningless. A bank running 12 times leverage and a software company with none cannot be ranked on this ratio.
- Goodwill distorts it. Acquisitive companies carry inflated book equity, and tangible common equity exists to correct for it.
- It ignores risk entirely. Two companies with identical ROE can carry completely different probabilities of not existing in five years.
- Negative equity makes it meaningless. A profitable company can report negative ROE, as McDonald’s does.
The Bottom Line
ROE asks a fair question: what return did the business generate on the capital shareholders have in it. The difficulty is that book equity is an accounting residue rather than a measure of capital, and it can be shrunk to nearly nothing, or below nothing, by a company doing something entirely ordinary.
None of this makes the ratio useless. For a company with stable financing, moderate leverage and no acquisition history, ROE tracked against its own past is genuinely informative about whether management is compounding shareholder capital. The failure mode is treating it as comparable across companies whose balance sheets differ, which is most of them.
Read it alongside return on assets, which uses a denominator that cannot be engineered away, and check the leverage that separates the two. Boeing’s 8.9% ROA in 2018 described a solid manufacturer. Its ROE above 3,000% described a treasury stock account, and only one of those was information about the company.
For the profit in the numerator see net income; for the repurchases that shrink the denominator see outstanding shares.
