Altman Z-Score Calculator

Three models, three answers

Edward Altman published the original Z-score in 1968, fitted on 66 listed American manufacturers, half of which had gone bankrupt. It worked well on that population and it has been applied ever since to populations it was never fitted on.

He himself published the revisions. Z′ refits every coefficient for private companies, because the original’s fourth term needs a market capitalisation and a private company has none. Z″ drops the sales-to-assets term altogether for non-manufacturers and emerging-market issuers, because asset turnover varies so much between service industries that no single coefficient fits.

The important part is that these are not the same model with different labels. The coefficients differ and so do the cut-offs. The original calls anything above 2.99 safe; Z′ uses 2.90; Z″ uses 2.60. A score of 2.7 is grey under the original, grey under Z′, and safe under Z″.

On the figures loaded in the tool above, the same balance sheet scores 3.15 under the original (safe), 2.43 under Z′ (grey) and 3.14 under Z″ (safe). Nothing about the company changed. Substituting book equity for market equity and rescaling the weights moved it a full zone.

That is why a Z-score quoted without its model is not a number anyone can check. It is also why running a private software business through the 1968 original — which several free calculators do — produces a figure with no defensible interpretation at all.

What the five ratios measure

Each term is a different kind of evidence, which is why the model beats any single ratio.

X1, working capital over assets is liquidity: how much of the asset base is short-term cover. It goes negative when current liabilities exceed current assets, which is exactly what the working capital calculator measures.

X2, retained earnings over assets is cumulative profitability, and by extension age. It is the term that does the most work; the next section is about why.

X3, EBIT over assets is current operating productivity — whether the assets are earning. It carries the largest coefficient in the original model, which is Altman saying that present earning power matters more than any single balance-sheet ratio.

X4, equity over liabilities is the solvency cushion: how far asset values could fall before liabilities exceed them. This is the term that differs most between models, because the original measures equity at market and the revisions at book.

X5, sales over assets is asset turnover. It is present in the original and Z′ and absent from Z″ entirely, because a consultancy and a steel mill have legitimately different turnover and no common weight can serve both.

Why retained earnings dominates the score

X2 carries a large coefficient in every model, and it is cumulative— the sum of every year of profit the company has ever retained, not this year’s.

That has a consequence people rarely anticipate. A company founded three years ago, profitable from the start and growing fast, has barely any retained earnings relative to its asset base simply because it has not existed long enough to accumulate them. It scores badly on X2 and its Z-score suffers accordingly.

This is not a defect in the model. In the population Altman fitted — mature listed manufacturers — a thin retained-earnings balance genuinely was a distress signal, because it meant a history of losses. In a population of young companies it means something entirely different, and the model cannot distinguish the two.

Anything that resets retained earnings has the same effect. A large one-off write-down, a restructuring charge, or a substantial dividend all reduce X2 without changing the operating business. A company that has bought back a lot of stock can even show negative retained earnings while trading perfectly well.

The tool shows each term’s contribution for exactly this reason. A low score driven by X2 on a young company is a different finding from a low score driven by X3, and only the breakdown distinguishes them.

Choosing the right model

Three questions settle it, in this order.

Is it listed? If there is no market capitalisation, the original model is simply unavailable — X4 has no numerator. Substituting book equity into the original formula is the single commonest error in Z-score calculators, because it silently applies the 1968 coefficients to a quantity they were never fitted against.

Is it a manufacturer? If not, the asset-turnover term is doing unreliable work and Z″ is the honest choice. A software business with almost no assets can post a large X5 that flatters its score for reasons unconnected to solvency.

Where is it domiciled? Z″ is also the model Altman recommends for emerging-market issuers, where accounting conventions and asset intensity differ from the fitted population.

When two models are arguably applicable, run both. If they agree the ambiguity did not matter; if they disagree, that disagreement is the finding and belongs in the write-up rather than being resolved by preference.

Where it struggles today

The model is more than fifty years old and it was fitted on a kind of company that is no longer typical. That does not make it useless, but it does bound where it applies.

Asset-light businesses break the denominators. Four of the five ratios divide by total assets. A company whose value is in brand, code or contracts carries a small asset base, which inflates X3 and X5 while X1 and X2 stay small. The score becomes unstable in a way that has nothing to do with distress risk.

Intangibles and goodwill distort the base. A company that grew by acquisition carries large goodwill in total assets, depressing every ratio that divides by it. An identical company that grew organically shows a better score for no economic reason.

Leases moved onto the balance sheet in 2019. IFRS 16 added right-of-use assets and lease liabilities to retailers and airlines, changing both total assets and total liabilities and therefore every term. Z-scores either side of that change are not comparable, the same way debt-to-equity is not.

Used as a screen — a way of ranking a portfolio or a supplier list for attention — it remains genuinely useful. Used as a verdict on one company, it was never that even in 1968.

What the Z-score cannot tell you

Three limits worth carrying.

It has no sense of timing. A distress-zone score is a statement about resemblance to companies that failed, not a prediction of when. Companies sit in the distress zone for years without failing, and some fail from the grey zone within months.

It cannot see the debt schedule. Nothing in the five ratios knows when repayments fall due. A company with a comfortable score and a bullet maturity next quarter is in more danger than the score suggests, which is what the DSCR calculator is for.

It is a screen, not a diagnosis. It was built to rank a population, and its published accuracy is a property of populations rather than of individual cases. A single score should prompt work, not conclude it.

Finally, it reads only what the accounts say. Fraud, undisclosed contingent liabilities and a concentrated customer base are all invisible to it, and all three have caused failures that no ratio model flagged.

Sources and methodology

The classification rules the inputs depend on.

• SEC Regulation S-X §210.5-02 — the balance-sheet captions behind total assets, total liabilities and retained earnings
• SEC Regulation S-X §210.5-03 — the income-statement format behind net sales and EBIT
• IAS 1 Presentation of Financial Statements — the current/non-current split that defines the working-capital term
• ACCA Financial Management technical articles — ratio-based failure prediction in the professional syllabus
• How we calculate — methodology

Method. All three models are implemented from one position rule with a per-model coefficient vector rather than as three separate formulas, so a coefficient cannot drift in one model without the suite noticing. The published coefficients and both thresholds for each model are asserted individually — wrong cut-offs would misclassify every company scored, and that failure would be silent. The disagreement the page is built on is itself a test case: the worked balance sheet must score safe under the original, grey under Z′ and safe under Z″, so if the models ever silently agreed the suite would fail rather than the page quietly becoming wrong. The engine is verified on every change against 87 assertions shared with the rest of the solvency cluster. The count and the per-case breakdown are published on the formula verification page.

Related calculators

The tests that look at the debt itself:

More in Business, or browse all calculators.