In its annual report for fiscal 2007, Lehman Brothers reported a return on average common stockholders’ equity of 20.8%. It was the firm’s headline profitability figure, down slightly from 23.4% the year before, and by that measure Lehman was performing well.
The same filing disclosed total assets of $691,063 million and net income of $4,192 million. That works out to a return on assets of about 0.61%. Lehman never printed the second number anywhere in the document. Nine months later it filed for Chapter 11.
One Ratio, One Substitution
The FDIC defines return on assets as net income, including gains or losses on securities and extraordinary items, as a percentage of average total assets, and calls it the basic measure of earnings performance. It defines return on equity identically except for the denominator: the same profit as a percentage of average total equity capital.
ROA = net income ÷ average total assets
That single substitution changes what the ratio can be made to say. Assets are what the business deploys. Equity is the sliver of those assets funded by shareholders rather than borrowed. A company can shrink the second without touching the first, and every unit of shrinkage inflates the equity return while leaving the asset return exactly where it was.
What Lehman’s Two Numbers Showed
Lehman’s own 10-K defines its leverage ratio as total assets divided by total stockholders’ equity. At November 30, 2007 that ratio stood at 30.7 times, up from 26.2 a year earlier.
Everything turns on the multiplier
Multiply a 0.61% return on assets by roughly 30 times leverage and you arrive in the high teens. The 20.8% was not evidence of exceptional business quality. It was a thin asset return amplified thirty-fold by borrowing, and the ratio that disclosed the amplification was the one the firm chose not to present.
The same leverage works in reverse. At 30 times, a loss of about 3.3% on assets erases equity entirely. An investor reading only the equity return had no way to see how little cushion sat beneath it; an investor comparing both ratios could calculate the leverage in a single division.
Any company reporting a strong equity return alongside a very low asset return is telling you that borrowing, not operating performance, produces most of the headline. The ratio between the two figures is the leverage multiple, and neither number alone reveals it.
Why Regulators Start Here
The FDIC’s Risk Management Manual of Examination Policies treats ROA as the starting point for earnings analysis, precisely because the denominator cannot be engineered away by financing decisions.
Its examiners are also instructed to treat an unusually high ROA as a warning sign rather than unambiguous good news. The ratio can be raised by buying riskier assets, and a bank earning far more per dollar of assets than its peers is generally taking more risk per dollar, not managing better. That is an unusual instruction, and it captures something most published commentary on profitability ratios misses: an outlier in either direction is a question, not a verdict.
The Gap Is Disclosed, When Anyone Looks
JPMorgan Chase publishes both figures side by side. Its 2025 annual report shows a return on common equity of 17%, a return on tangible common equity of 20%, and a return on assets of 1.29%.
The bank closed the year with $4,424,900 million of total assets against $362,438 million of total stockholders’ equity, about 12.2 times leverage.
Nothing is concealed. Both ratios sit in one table, and the 13-fold gap between them is a straightforward consequence of running a balance sheet at twelve times equity. What makes Lehman instructive is not that it hid the leverage, which it disclosed at 30.7 times, but that it presented only the flattering half of the pair.
The industry pattern
Across FDIC-insured institutions, ROE ran at roughly 9.6 to 10.9 times ROA every year from 2002 to 2008. In 2006 the industry earned 1.28% on assets and 12.30% on equity; in 2003, 1.38% and 15.05%. The relationship is arithmetic, and stable, until leverage changes.
Why the ratio between them stays steady
Across those years ROE ran at roughly 9.6 to 10.9 times ROA, and the consistency is not a coincidence. Bank leverage is constrained by capital requirements, so the multiplier connecting the two ratios sits within a regulated band.
When that multiple moves sharply at a single institution, it is telling you the balance sheet changed, not that the business improved.
Lehman, as an investment bank rather than a deposit-taking institution, was not subject to the same constraint. Its 30.7 times sat roughly three times above the commercial banking norm of the period, and the equity return it produced was correspondingly inflated.
The most recent published quarter shows FDIC-insured institutions reporting an ROA of 1.24% on aggregate net income of $77.7 billion in the fourth quarter of 2025.
ROA Is Harder to Flatter, Not Impossible
Because assets rather than equity sit in the denominator, buybacks do not move ROA at all. A company can repurchase shares until its book equity is negative and its return on assets will be unchanged, and the ratio survives situations that break ROE.
McDonald’s illustrates this. It closed 2025 with a shareholders’ equity deficit of $(1,791) million on $59,515 million of total assets, following $79,316 million of cumulative treasury stock repurchases, while earning $8,563 million of net income. Return on equity is meaningless for that company. Return on assets is perfectly computable, and McDonald’s itself reports an after-tax return on invested capital of 20.3% rather than an equity return.
The contrast with ROE is instructive here. Both companies had spent years returning capital, and both ended with book equity that no longer described anything useful. ROE registered that as a broken ratio. ROA registered it as nothing at all, because the assets were still there, still generating profit, and still in the denominator where they belonged.
The ratio still has soft spots. Asset values are historical costs less depreciation, so an old asset base understates the denominator and flatters the return. Acquisitions add goodwill, inflating assets and depressing ROA without any operational change. And asset-light businesses will always post higher ROA than capital-intensive ones, which makes cross-sector comparison meaningless.
Computing It From a Filing
Both inputs are disclosed, and the calculation takes a minute. Net income is the income statement's final line. Total assets is the subtotal on the balance sheet. Averaging the opening and closing asset figures gives the denominator regulators use, though the year-end figure is close enough for most purposes unless the balance sheet moved sharply.
Two cautions apply. Use the same profit definition consistently, since a company quoting adjusted earnings will produce a higher ROA than the GAAP figure would. And be careful with companies that made a large acquisition mid-year: the assets arrive on the balance sheet immediately while the earnings they generate appear only for the remaining months, which depresses the ratio for one period without anything being wrong.
For a bank, the figure is usually published directly, since regulators require it. For an industrial company it typically is not, so it is worth computing rather than waiting for it to be presented.
What ROA Cannot Tell You
- Cross-industry comparison does not work. A software company and a utility differ by an order of magnitude on this ratio for structural reasons.
- Historical cost distorts the denominator. Long-held, heavily depreciated assets make the return look better than a like-for-like comparison would.
- A high reading can mean risk. FDIC examiners are told to treat an outlier as a possible sign of aggressive asset selection.
- It inherits the numerator’s problems. Net income carries one-time items and non-cash gains, so a distorted profit figure produces a distorted ROA. Read it against free cash flow.
- It says nothing about how the assets were funded. That is the question return on equity answers, and the pair is more informative than either alone.
Divide reported ROE by reported ROA. The result is roughly the leverage multiple, and it takes one calculation. For Lehman in 2007 the answer was about 34; for JPMorgan in 2025 it is about 13; for an unlevered industrial company it will be close to 1.5. That single number frames every other profitability figure in the accounts.
Reading the Pair Together
Neither ratio is the better one. They answer adjacent questions, and the value is in the relationship.
ROA asks whether the business converts the assets it controls into profit. ROE asks what shareholders earn on the portion they funded. When both are healthy and the gap between them is modest, the company is generating genuine returns without much borrowing. When ROE is strong and ROA is thin, the return is manufactured on the balance sheet, and it will reverse with the same multiplier in a bad year.
ROA also degrades gracefully where ROE does not. Industry ROA turned negative in the crisis, reaching -0.94% in the final quarter of 2008, the worst since -1.10% in 1987's second quarter. A negative ROA means the assets lost money, which is unambiguous. A negative ROE can mean the same thing or can mean the denominator went below zero, and the reader cannot tell which without checking.
The Bottom Line
Return on assets is the less flattering of the two profitability ratios and the more reliable. It cannot be improved by buybacks, it survives negative book equity, and regulators use it as the starting point for earnings analysis for exactly those reasons.
Lehman’s 20.8% and its 0.61% were both true in the same document. One was published and one had to be computed, and the gap between them was the leverage that ended the firm. Whenever you are handed an equity return, work out the asset return underneath it before deciding what the first number meant.
For the amplified version of this ratio see return on equity; for the profit figure both share see net income.
