What the Altman Z-Score estimates
Edward Altman's 1968 paper asked whether bankruptcy within two years could be predicted from published accounts. His method was discriminant analysis over a sample of manufacturers, and the output was a formula: five ratios, each multiplied by a coefficient the analysis produced, summed into one score. The higher the score, the further the company sits from the pattern that preceded failure in his sample.
That is a narrower claim than it is usually given credit for. The Altman Z-Score is a distance-from-distress estimate calibrated on a particular kind of company in a particular period. It is not a valuation, not a credit rating, and not a probability, though it correlates with all three well enough to have survived more than fifty years of use.
The Altman Z-Score formula and its calculation, term by term
The Altman Z-Score formula for public manufacturers is Z = 1.2·A + 1.4·B + 3.3·C + 0.6·D + 1.0·E. A is working capital over total assets, a liquidity measure. B is retained earnings over total assets, which captures cumulative profitability and, indirectly, age. C is earnings before interest and tax over total assets: operating productivity, and the term with the largest coefficient. D is the market value of equity over the book value of total liabilities, the only term that uses a market price. E is sales over total assets, asset turnover.
The coefficients in the Altman Z-Score formula are the finding, not a choice you get to make. C carries 3.3 because operating profitability did most of the separating in the original sample; E carries 1.0 because turnover did least. Anyone adjusting the weights has built a different model, which may well be better, and should not call the result an Altman Z-Score.
The term that changes dailyTerm D uses the market value of equity, so the Altman Z-Score of a listed company moves whenever the share price moves, with no change in any filing. A score quoted without a date is therefore incomplete, and a score that fell sharply may reflect nothing except a re-rating of the stock.
The three zones, and what they claim
Altman's published bands for this version are: below 1.81 is the distress zone, above 2.99 is the safe zone, and between them is the grey zone where the model does not discriminate. The grey zone is not a hedge; it is a stated finding that a substantial band of scores carried no reliable signal, and reporting it as though the answer were "mildly risky" misrepresents the calculation.
It also matters that these are zones rather than thresholds with meaning at the boundary. A company at 1.79 and one at 1.83 are not meaningfully different; the formula's inputs are annual accounting figures with their own tolerances. Precision to two decimal places is arithmetic precision, not measurement precision.
A worked calculation, all five terms
Take a listed manufacturer with total assets of 1,250 and total liabilities of 480. Working capital is 120, retained earnings 340, EBIT 95, sales 1,100, and the market values its equity at 900. The five ratios are therefore 120/1250 = 0.096, 340/1250 = 0.272, 95/1250 = 0.076, 900/480 = 1.875 and 1100/1250 = 0.880.
Applying the published coefficients: 1.2 × 0.096 = 0.115, plus 1.4 × 0.272 = 0.381, plus 3.3 × 0.076 = 0.251, plus 0.6 × 1.875 = 1.125, plus 1.0 × 0.880 = 0.880. The sum is 2.75, which lands in the grey zone: above the 1.81 distress threshold and below the 2.99 safe one, in the band the original analysis found no reliable discrimination in.
The decomposition is the useful part. Two terms carry three quarters of that score: the equity-to-liabilities term at 1.125 and asset turnover at 0.880. The operating profitability term, which has the largest coefficient of all five, contributes only 0.251 because the underlying ratio is small. This company is scoring on its market valuation and its sales volume rather than on what it earns. A one-third fall in the share price would take the equity term from 1.125 to 0.75 and the total from 2.75 to 2.38. Still inside the grey zone, but a sixth of the score gone with nothing whatever having changed in the accounts. Reaching the distress threshold would take a fall of roughly 84 percent, which is the honest measure of how much of this particular score is a market opinion rather than an accounting fact.
The variants, and why they exist
The original formula needs a market price, which private companies do not have. Altman published a Z'-score for them, replacing term D with book equity over total liabilities and re-fitting every coefficient: Z' = 0.717·A + 0.847·B + 3.107·C + 0.420·D + 0.998·E, with distress below 1.23 and safe above 2.90. There is also a Z''-score for non-manufacturers and emerging markets that drops the turnover term entirely, because asset turnover is not comparable across service businesses.
The existence of three sets of coefficients is the clearest statement of the model's scope. Applying the manufacturing formula to a software company or a bank produces a number, and that number is outside the conditions the coefficients were fitted under. Choosing the right variant is not a refinement; it is a precondition for the calculation meaning anything.
What the score cannot see
Every input is a balance-sheet or income total, which means the model has no view on terms. A company with a comfortable score can have its entire debt balance falling due in eighteen months; nothing in the five ratios asks when. It cannot see covenants either, so a business that will breach a leverage test next quarter and have its borrowing accelerated looks exactly like one that will not.
It is also blind to what sits outside the reporting entity, and to concentration of any kind: one customer, one supplier, one product. Those are the failure modes that arrive quickly, and they live in the notes rather than in the totals. The Z-Score is a useful reading of the shape of a balance sheet, and reading the shape is not the same as reading the timetable.
Why a 1968 model is still in use
The obvious objection to the Altman Z-Score is its age: the coefficients were fitted on a sample of manufacturers more than fifty years ago, on accounting standards that have since changed substantially. That objection is correct and the model is still used, and both facts have the same explanation. It is transparent. Every input is a published figure, every weight is stated, and any user can see exactly which term moved the result, which is more than can be said for most of what has replaced it.
Its practical role has shifted accordingly. It is no longer treated as a bankruptcy prediction so much as a standardized way of asking whether a balance sheet looks like the ones that failed, and of decomposing the answer. That is a modest claim, it is the claim the original paper made, and a model that has been honest about its own scope for five decades is a rarer thing than an accurate one.
Using the Z-Score in screening
As a screening criterion the Altman Z-Score is most useful as an exclusion rather than a ranking. Screening out the distress zone removes a specific, identifiable failure pattern from a shortlist. Ranking by the score rewards asset-light, highly profitable companies for reasons that are partly artefacts of the ratios rather than judgements about their prospects.
The other practical use is decomposition. Because the formula is a weighted sum, you can see which term is dragging a score down: thin working capital, negative retained earnings, weak operating profitability. That points directly at the part of the accounts worth reading. The calculator here shows each term's contribution for exactly that reason, rather than only the total.