← Quantery Blog

ROE and ROA: taking return on equity apart

September 12, 2026 · 8 min readfundamentalsprofitability

Return on equity tells you how much net income a business earned against the capital belonging to its shareholders. Return on assets asks how much it earned against the whole asset base. Read together, they separate a productive business from one whose return to owners has been enlarged by a thin equity cushion.

ROE can rise because margins improved, assets worked harder, or leverage increased. Leverage is the use of debt to fund assets. The DuPont decomposition puts those routes on separate lines: net margin times asset turnover times the equity multiplier. If you don't take ROE apart, a fine-looking percentage can hide the part doing the work.

This builds on the balance-sheet identity, where equity is the residual after liabilities, and the income statement from revenue to net income. Those two statements meet inside both ratios.

What do ROE and ROA measure?

The standard formulas use net income from a period and average balance-sheet values over that period:

ROE = net income / average shareholders' equity
ROA = net income / average total assets

Return on equity (ROE) measures profit against the shareholders' accounting claim. A positive ROE says the business earned money on that claim. A higher ROE says more profit was produced per dollar of equity, but it doesn't tell you how the result was achieved.

Return on assets (ROA) measures the same profit against everything the company controls: cash, inventory, factories, receivables and acquired assets. The denominator includes assets funded by creditors as well as assets funded by shareholders. That makes ROA less sensitive to the funding mix than ROE.

Use averages because net income is a flow earned across a period while equity and assets are snapshots on particular dates. The usual approximation is the opening balance plus the closing balance, divided by two. This plain-language treatment of profitability ratios uses average assets for ROA and average equity for ROE. Dividing a full year of income by the final day's balance can distort the ratio after a large acquisition, disposal or buyback.

There's a second limit hiding in the numerator. Net income sits after interest expense, so ROA isn't fully independent of financing. More debt can bring more assets into the business, and its interest cost can reduce the earnings those assets leave behind. ROA removes the shrinking-equity trick from the denominator. It doesn't remove debt from the company.

Why can leverage make ROE look stronger?

Start with the balance-sheet equation:

assets = liabilities + equity

Now imagine the same operating assets supported by less equity and more debt. If earnings hold, the ROA numerator and denominator barely move. The ROE denominator gets smaller, so ROE rises. Shareholders have a narrower claim supporting the same asset base, and each dollar of that claim appears to earn more.

That arithmetic cuts both ways. Interest has to be paid before anything reaches common shareholders. A small decline in asset earnings lands on a smaller equity cushion, so losses are magnified too. ROE rewards the upside of leverage in a good period without printing the downside beside it.

Buybacks can produce a related effect. Cash leaves the asset side and equity falls by the same amount. If later net income stays level, ROE rises even though the underlying business didn't become more productive. The balance-sheet article used McDonald's to show why sustained distributions can even push healthy-company equity below zero. At that point ROE stops being a useful percentage. A positive numerator divided by negative equity is an accounting oddity, not evidence of terrible operations.

This is why a thesis shouldn't rank on ROE alone. Pair it with ROA, then put leverage on the row where you can see it. A high ROE beside a decent ROA is a claim about operations. A high ROE beside a thin ROA and a large asset-to-equity multiple is a claim about funding.

Lehman's ROE came with a very small cushion

Lehman Brothers' fiscal 2007 filing is a clean historical example because it printed an attractive shareholder return and the balance sheet needed to produce it on the same document.

For the year ended November 30, 2007, Lehman's Form 10-K reported net income of $4.192 billion and return on average common shareholders' equity of 20.8 percent. At year-end it carried $691.063 billion of assets against $22.490 billion of stockholders' equity. Using the filing's opening and closing asset balances gives an approximate ROA of 0.7 percent.

The Financial Crisis Inquiry Commission described that year-end position as leverage of 31 to 1. It also found that the firm's failure came from concentrated illiquid real-estate positions, excessive risk and controls that management breached or ignored. Lehman filed for bankruptcy on September 15, 2008.

ROE didn't cause any of that, and a low ROA wouldn't have predicted the exact path. The pair showed something narrower and still useful. A return above twenty percent for common owners rested on less than one percent earned across the asset base, amplified through a very small equity slice. Anyone reading the ROE by itself saw the output and missed the mechanism.

Financial firms deserve their own ratios because borrowing and financial assets are the machinery of the business. Quantery's Buffett Quality Value and Piotroski F-Score templates exclude Financial Services and Real Estate for that reason. Lehman belongs here to explain leverage. It doesn't belong in a general industrial-company rule.

DuPont analysis shows where ROE came from

The DuPont decomposition rewrites ROE as three linked ratios:

ROE = net margin × asset turnover × equity multiplier

net margin        = net income / revenue
asset turnover    = revenue / average assets
equity multiplier = average assets / average equity

Multiply them and revenue cancels, then average assets cancels. You're left with net income over average equity, which is ROE. The arithmetic hasn't changed. The explanation has.

Net margin asks how much of each sales dollar reached net income. It captures pricing, direct costs and everything else that sits between the top and bottom lines.

Asset turnover asks how much revenue each dollar of assets produced. A low-margin distributor can earn a respectable return by moving a great deal of product through a modest asset base. An asset-light software business may travel the other route, with slower turnover and much more profit left from each sale.

The equity multiplier asks how many dollars of assets sit above each dollar of shareholder equity. An answer close to one means owners fund most of the assets. As the multiple grows, liabilities are doing more of the funding. It isn't a verdict, but it tells you how much of ROE came from leverage.

This decomposition also stops bad comparisons. Two companies can post the same ROE with opposite economics. One may run on thin margins and fast inventory movement. Another may carry fat margins while its assets turn slowly. Add a high equity multiplier and you've found a third route. The headline ratio treats them as equals. The components don't.

How should a Quantery thesis calculate them?

Match the periods before setting a threshold. Use trailing net income and revenue, then average the current balance sheet with the same fiscal quarter from the prior year. The article on trailing versus forward numbers explains why lag(field, 4) targets that year-over-year comparison. Because lag is positional in the Quantery DSL (four filed periods back), a filing gap shifts which period you land on; add a history_periods >= 5 guard or treat the average as null when either value is missing.

The shape below is illustrative. The bundled templates and Quantery DSL reference are the reference for exact fields. This version also refuses to divide by non-positive equity, because a giant negative ROE isn't a useful ranking signal.

# "Strong returns, with the leverage visible" - illustrative
params:
  roe_strong:     0.15
  roa_strong:     0.08
  multiplier_max: 4

features:
  ni_ttm:       ttm(net_income)
  rev_ttm:      ttm(revenue)
  assets_now:   newest(total_liabilities) + newest(total_equity)
  assets_prior: lag(total_liabilities, 4) + lag(total_equity, 4)
  equity_now:   newest(total_equity)
  equity_prior: lag(total_equity, 4)
  avg_assets:   if(assets_now > 0 and assets_prior > 0,
                   (assets_now + assets_prior) / 2, null)
  avg_equity:   if(equity_now > 0 and equity_prior > 0,
                   (equity_now + equity_prior) / 2, null)
  roa:          if(avg_assets > 0, ni_ttm / avg_assets, null)
  roe:          if(avg_equity > 0, ni_ttm / avg_equity, null)
  net_margin:   if(rev_ttm > 0, ni_ttm / rev_ttm, null)
  asset_turnover: if(avg_assets > 0, rev_ttm / avg_assets, null)
  equity_multiplier: if(avg_equity > 0,
                         avg_assets / avg_equity, null)
  dupont_roe: net_margin * asset_turnover * equity_multiplier

criteria:
  returns:
    rules:
      - { when: "is_null(roe) or is_null(roa)", score: 0, flag: no_return_base }
      - { when: "roe >= $roe_strong and roa >= $roa_strong", score: 2 }
      - { when: "roe > 0 and roa > 0", score: 1 }
      - { else: 0 }
  leverage:
    rules:
      - { when: "is_null(equity_multiplier)", score: 0, flag: no_multiplier }
      - { when: "equity_multiplier <= $multiplier_max", score: 2 }
      - { else: 0 }

gate:
  mode: strict

The null guards are doing real work. Because lag is positional, a filing gap can shift which quarter you land on; when either prior balance is missing the average resolves to null, and the company can't collect points from a denominator assembled across mismatched filings. If equity is zero or negative, the return and multiplier are null. The result says this lens can't measure the company, which is better than a dramatic ratio with no economic meaning.

The thresholds are where your thesis begins. Run one version that scores ROE alone. Run another that requires ROA as well, then cap the equity multiplier. Backtest them over the same universe and rebalance schedule. The results are hypothetical and exclude costs, so you're comparing rules instead of estimating an account balance. If the apparent edge disappears when leverage is constrained, you learned what the first rule was rewarding.

ROE is useful because it compresses a lot of business into one number. That's also its defect. Keep the compression for ranking, keep the components beside it for diagnosis, and make every threshold yours to move and test.

Want to try this on your own rules? Quantery is free for 14 days: the full app, no card required.

← All articles