Altman Z-Score, Explained: Reading Bankruptcy Risk With the Exact Formula

The Altman Z-Score compresses five balance-sheet and market ratios into a single number that flags financial distress. Here is the exact formula, what each ratio measures, the zones, the variants, and where the model quietly stops working.

The five ratios that build the score

The original Altman Z-Score compresses five financial ratios into a single number using coefficients Edward Altman derived from multiple discriminant analysis in 1968. Each ratio is a lens on a different dimension of corporate health, and each is deliberately scaled by total assets or total liabilities so that a small company and a large one can be compared on the same axis. For a public manufacturer the formula is:

Z = 1.2·X1 + 1.4·X2 + 3.3·X3 + 0.6·X4 + 1.0·X5

• X1 — Working Capital / Total Assets. Short-term liquidity relative to size. Working capital is current assets minus current liabilities. A firm drawing down its liquidity cushion sees this ratio shrink, and it can turn negative when current liabilities exceed current assets.
• X2 — Retained Earnings / Total Assets. Cumulative profitability and, indirectly, age. Retained earnings is the stack of profits a firm has kept rather than distributed over its entire life, so a young company or one that has absorbed large losses carries a low or negative X2 even if the current year is healthy.
• X3 — EBIT / Total Assets. The operating productivity of assets, before the distortions of leverage and taxes. Altman weighted this most heavily (3.3) because the capacity of assets to generate operating income is the core of solvency.
• X4 — Market Value of Equity / Total Liabilities. How far the firm's market value can fall before liabilities exceed assets. This is the only ratio that pulls in a market price rather than an accounting figure, which is why it needs different treatment for private firms.
• X5 — Sales / Total Assets. Asset turnover — how much revenue each dollar of assets produces. This ratio is highly industry-dependent, which becomes important when we reach the variants.

A full worked calculation

Numbers make the mechanics concrete. Consider Illustrative Manufacturing Co., a hypothetical public manufacturer with the figures below (in millions of dollars). These are round numbers chosen for teaching, not a real company.

Computing each ratio and applying its coefficient:

• X1 = 300 / 1,000 = 0.30 → 1.2 × 0.30 = 0.360
• X2 = 250 / 1,000 = 0.25 → 1.4 × 0.25 = 0.350
• X3 = 120 / 1,000 = 0.12 → 3.3 × 0.12 = 0.396
• X4 = 800 / 500 = 1.60 → 0.6 × 1.60 = 0.960
• X5 = 900 / 1,000 = 0.90 → 1.0 × 0.90 = 0.900

Summing gives 0.360 + 0.350 + 0.396 + 0.960 + 0.900 = 2.966, or about 2.97. That lands just inside the grey zone, a hair below the 2.99 safe threshold. Notice how much of the total comes from X4 and X5 — a solid equity cushion and healthy asset turnover carried this firm, while its EBIT contribution was modest. Changing one input shows the sensitivity: if the market value of equity fell from 800 to 400, X4 would drop to 0.80, its weighted contribution to 0.48, and the total to roughly 2.49 — deeper into grey territory — without a single operating number changing. That single-ratio leverage is why the score is read as a composite, never as any one line item.

Z, Z-prime, and Z-double-prime

The original Z-Score has two well-known descendants because two of its assumptions do not travel well. X4 needs a market price, which private companies do not have, and X5 embeds an industry bias that penalizes asset-heavy businesses and flatters asset-light ones.

Z′ (private manufacturers) swaps the market value of equity in X4 for book value of equity and re-estimates every coefficient:

Z′ = 0.717·X1 + 0.847·X2 + 3.107·X3 + 0.420·X4 + 0.998·X5

where X4 = Book Value of Equity / Total Liabilities. Its bands sit lower than the original: above 2.90 is safe, 1.23 to 2.90 is grey, and below 1.23 is distress.

Z″ (non-manufacturers and emerging markets) goes further and drops X5 entirely, removing the asset-turnover ratio that varies most across industries. That makes it appropriate for retailers, service firms, and other businesses whose asset bases are not comparable to a factory's:

Z″ = 6.56·X1 + 3.26·X2 + 6.72·X3 + 1.05·X4

again with book value of equity in X4. Its bands are: above 2.60 safe, 1.10 to 2.60 grey, below 1.10 distress. A common variant adds a constant of 3.25 to produce an "emerging-market score" on a comparable scale, but the four-variable model above is the standard form.

The three variants sit side by side like this:

To see Z″ in action, take Illustrative Retail Co., a hypothetical non-manufacturer with working capital of 150, total assets of 1,000, retained earnings of 300, EBIT of 90, book value of equity of 400, and total liabilities of 600 (all in millions):

• X1 = 150 / 1,000 = 0.15 → 6.56 × 0.15 = 0.984
• X2 = 300 / 1,000 = 0.30 → 3.26 × 0.30 = 0.978
• X3 = 90 / 1,000 = 0.09 → 6.72 × 0.09 = 0.605
• X4 = 400 / 600 = 0.667 → 1.05 × 0.667 = 0.700

The total is 0.984 + 0.978 + 0.605 + 0.700 = 3.267, about 3.27 — comfortably above the 2.60 safe threshold for Z″. The much larger coefficients here (6.56 versus 1.2) are not a sign this firm is healthier than the manufacturer above; the variants live on different scales, and their scores are not interchangeable.

Reading the zones — and the trend

For the original Z, a score above 2.99 sits in the safe zone, between 1.81 and 2.99 in the grey zone, and below 1.81 in the distress zone. The grey zone is not a verdict; it is an admission of uncertainty — historically many firms passed through it in both directions. What often carries more information than any single reading is the direction of travel.

Suppose a manufacturer posts scores of 3.4, 2.8, 2.1, and 1.6 across four consecutive years. Each figure describes that year's balance sheet, but the sequence tells a fuller story: a firm that began comfortably safe has crossed the whole grey zone and moved into distress territory, and the roughly half-point annual decline is itself worth understanding. A single 1.6 could be industry noise; a steady glide from 3.4 to 1.6 is a pattern. Investors often examine the trajectory and, crucially, the ratio driving it — a falling X3 (operating income shrinking against assets) points somewhere very different from a falling X4 (the market re-rating the equity downward). The score is descriptive of stress, and its slope frequently says more than its level.

Accuracy, history, and where it breaks

Altman built the model on 66 companies — 33 manufacturers that went bankrupt between 1946 and 1965 and 33 that survived. In that original sample it correctly classified roughly 94% of the bankrupt firms one year before failure and about 72% two years before. Out-of-sample studies over the following decades generally reported accuracy in the 80–90% range one year ahead, which is durable performance for a formula that has never been re-fitted in its classic form.

Two error types matter. A Type I error labels a firm safe when it later fails; a Type II error labels a healthy firm distressed. The model is tuned to worry more about missing a real failure, so false alarms — sound companies scoring in the grey or distress zone for structural reasons — are common and expected. Known failure modes include:

• Financial firms. Banks, insurers, and other lenders carry leverage that would be alarming anywhere else as a normal feature of their business, and their balance sheets are not built from the working-capital and inventory logic the ratios assume. The score is not meaningful for them.
• Young and asset-light companies. A startup with accumulated losses has a deeply negative X2, and a software firm with few tangible assets can produce distorted X3 and X5 readings, pushing genuinely viable businesses into low scores.
• Accounting choices and off-balance-sheet items. The score reads only what the statements report. Leases, contingent liabilities, or aggressive revenue timing can flatter the inputs without changing the underlying risk.
• Point-in-time snapshots. A single reading misses seasonality and one-off items; the model was validated on annual data, not one unusual quarter.

A low score is best read as balance-sheet stress relative to a 1960s manufacturing benchmark — a prompt to look closer, not a prediction that a company will fail.

How it fits with Ohlson O, Merton, and Piotroski

No single distress model is authoritative, and each was built on different mathematics, so they fail in different places — which is precisely why analysts often read them together rather than in isolation.

• Ohlson O-Score (1980) uses logistic regression across nine variables and outputs a probability rather than a zone, which some find easier to interpret than a raw score. It leans on size, leverage, and liquidity and does not require a market price.
• Merton distance-to-default (1974) is structural and market-based: it treats a company's equity as a call option on its assets and asks how many standard deviations of asset value sit between the firm and the point where liabilities exceed assets. Because it uses live market prices and volatility, it reacts faster than any accounting model — and, for the same reason, it inherits the market's mood swings.
• Piotroski F-Score (2000) is not a bankruptcy model at all; it is nine binary tests of fundamental strength and year-over-year improvement, scored 0 to 9. It answers "is this balance sheet getting healthier?" rather than "is this firm near default?"

Read together they triangulate: an accounting snapshot (Z), a statistical probability (O), a market-implied view (Merton), and a fundamental-momentum check (Piotroski). When they agree, the picture is consistent; when they disagree, the disagreement itself is informative.

Common mistakes and edge cases

1. Using the original Z on a private company. Without a market cap there is no X4, so the classic formula forces a book value into a coefficient calibrated for market value. Use Z′ instead.
2. Scoring a bank or insurer. The most frequent misuse. Financial-sector leverage breaks every assumption in the model, and the number it produces is not comparable to an industrial firm's.
3. Confusing book and market value in X4. A common driver of a wrong original-Z is plugging book equity where market value belongs, which typically understates the score for firms trading above book value.
4. Comparing scores across variants. A Z″ of 3.3 and a Z of 2.9 are not on the same ruler. Coefficients and thresholds differ, so a firm should be tracked with one consistent variant over time.
5. Reading one snapshot in isolation. A grey-zone reading after a bad quarter, or a distress reading for a young growth firm, is often a structural artifact rather than a warning. The trend and the driving ratio matter more than the level.

Frequently asked questions

What is a good Z-Score?

For the original Z used on public manufacturers, a reading above 2.99 has historically been associated with the safe zone. For Z′ the safe threshold is 2.90 and for Z″ it is 2.60. "Good" is relative to the variant and the industry, and a high score describes balance-sheet resilience rather than any prediction about the share price.

Why doesn't the Z-Score work for banks?

The ratios assume an industrial balance sheet built from working capital, inventory, and operating assets. Banks and insurers run high leverage as a normal condition of their business and structure their assets and liabilities entirely differently, so coefficients calibrated on 1960s manufacturers produce a number with no reliable meaning for them.

Z or Z-double-prime — which one do I use?

Match the variant to the company. Public manufacturers fit the original Z; private manufacturers fit Z′; non-manufacturers such as retailers, service businesses, and many emerging-market firms fit Z″, which drops the industry-sensitive sales-to-assets ratio. Whichever you choose, use it consistently so scores stay comparable across periods.

Does a low Z-Score mean a company will go bankrupt?

No. A low score indicates balance-sheet stress relative to the model's benchmark, and false alarms are common — the model is deliberately tuned to flag rather than miss potential trouble. Many low-scoring firms restructure, recover, or were simply mis-scored because of their industry or age. It is one input among many, not a forecast.

How often should the Z-Score be recomputed?

Because the inputs come from financial statements, the score naturally updates each time a company reports — quarterly or annually. The X4 term, which uses market value in the original version, drifts continuously with the share price, so that component can be refreshed as often as the market moves. Watching the score across several reporting periods is generally more informative than any single calculation.

How Quintarthai helps

Quintarthai computes the Altman Z-Score and its Z′ and Z″ variants deterministically from public filings (SEDAR+/EDGAR/SEDI) and licensed market data for US and Canadian equities, selecting the appropriate variant, applying bank and not-meaningful gates where the model does not apply, and displaying the underlying five ratios alongside the score so the figure can be traced back to its inputs. It presents this together with related distress and quality models — including Ohlson O, Merton distance-to-default, and Piotroski F — as educational reference data, not investment advice.