Altman Z-Score: Formula and Limits
Table of Contents
The Altman Z-Score combines several financial ratios into a single screen for corporate financial distress. It can help you decide which company’s accounts deserve closer investigation, but only if you select the appropriate model, use consistent inputs and interpret the result within its historical limits.
The most consequential mistakes happen before the arithmetic: using book equity in a market-equity formula, applying manufacturing cutoffs to a service-company score, or reading the answer as a current probability of bankruptcy. This guide explains the original Z, private-company Z-prime and nonmanufacturer Z-double-prime models, then works through financial statements and the questions a score should prompt.
What Is the Altman Z-Score?
Edward Altman’s model uses multiple discriminant analysis: a statistical method that finds a weighted combination of characteristics that separates groups. Published in 1968, the original model compared manufacturers that became bankrupt with manufacturers that did not. Its output is a score indicating how closely a firm’s financial profile resembles those groups. [1]
A lower score generally signals a weaker financial profile within the chosen model. It does not establish that bankruptcy will occur, when it will occur, or the percentage chance that it will occur.
The approach extends ordinary financial ratio analysis by assigning fixed weights to several measures together. A profitable company can still have weak liquidity or substantial liabilities; combining dimensions helps organize that conflicting evidence. The fixed weights also impose a limitation: they were estimated for particular samples rather than tailored to every business.
This article uses U.S. nonfinancial companies as its working context. Banks and insurers require separate sector analysis; the label “nonmanufacturer” does not make these formulas universally suitable for financial institutions.
The Financial Statement Inputs
Define the inputs before choosing the equation. Let TA mean total assets and TL mean the book value of total liabilities. The first three ratios are shared across the models below. The fourth changes its equity definition, and the fifth disappears from Z-double-prime. [2]
| Input | Definition | Economic meaning |
|---|---|---|
| X1 | (Current assets − current liabilities) / TA | Net working capital relative to the asset base |
| X2 | Retained earnings / TA | Accumulated profitability retained in the business |
| X3 | Earnings before interest and taxes / TA | Earning power before financing costs and income taxes |
| X4,M | Market value of equity / TL | Market-valued equity cushion; used in original Z |
| X4,B | Book value of equity / TL | Accounting equity cushion; used in Z-prime and Z-double-prime |
| X5 | Annual net sales / TA | Asset turnover; used in Z and Z-prime |
Working capital is a dollar amount, not the current ratio. Subtract current liabilities from current assets before dividing by total assets. The working capital management framework helps explain why receivable collection, inventory and supplier terms affect that amount.
Retained earnings are accumulated earnings after distributions and other relevant accounting adjustments, rather than this year’s net income or cash balance. A company can report a retained deficit, producing a negative X2. EBIT can also be negative. Preserve those signs rather than replacing losses with zero.
Market equity is the combined market value of common and preferred shares, where applicable. Common market capitalization alone is sufficient only when there is no preferred equity to include. Book equity comes from the balance sheet. Neither equity measure is enterprise value.
Total liabilities include current and long-term liabilities, including obligations such as accounts payable; they are broader than interest-bearing borrowings. Therefore X4,B is not generally the reciprocal of a debt-to-equity ratio that counts only borrowing as debt. The SEC’s financial statement guide distinguishes these balance-sheet categories. [3]
Use Consistent Units and Dates
Enter each ratio as a decimal or multiple in the equations below: use 0.10 for 10% and 1.20 for sales equal to 1.2 times assets. Use the same currency and scale for each numerator and denominator. A final Z-score is dimensionless; it is not expressed as a percentage.
Use annual sales and annual EBIT with the relevant year-end balance-sheet values for this historical-model calculation. Record the market-equity date too. If you instead use trailing twelve-month earnings with a recent interim balance sheet, disclose that choice and compare like with like. Do not silently replace year-end assets with average assets just because another profitability ratio uses averages.
Document any asset adjustments in the implementation you are following. Altman’s later discussion specifies tangible assets in its input discussion, while the displayed equations are labeled total assets. The worked example below assumes no intangible assets, so reported total assets and tangible assets coincide. Large goodwill balances or inconsistent vendor adjustments need investigation rather than an unexplained denominator swap. [2]
Choose the Correct Altman Formula
The model choice depends on the company population and available equity data. Choose it before calculating. Switching formulas to obtain a more comfortable result defeats the purpose of the screen.
Original Z: Publicly Traded Manufacturers
This is the familiar public-manufacturer formulation presented in Altman’s later author discussion. Some presentations retain 0.999 rather than 1.0 for the sales coefficient, producing a small numerical difference. Older presentations also use smaller coefficients with differently scaled inputs. Keep the coefficient convention and input units together; do not combine a percentage-scaled equation with decimal-scaled ratios. [2]
Z-Prime: Private Manufacturers
A privately held manufacturer lacks a directly quoted market value for its equity. Z-prime, written Z′, uses book equity and reestimates all five coefficients. It is not the original equation with just one numerator replaced. [2]
Z-Double-Prime: Nonfinancial Nonmanufacturers
Z-double-prime, written Z″, removes sales/assets because asset turnover varies substantially across industries, then changes the remaining weights. It uses book equity and is a common starting version for nonfinancial service and other nonmanufacturing firms, including public or private firms. Industry suitability still needs assessment. [2]
Altman’s emerging-market scoring and bond-rating-equivalent work also uses an adaptation that adds 3.25. That is a different calibration. Do not add 3.25 to the displayed no-constant equation and then apply the no-constant zones below. [2]
How to Interpret the Z-Score Zones
Historical zone boundaries differ by model. The table uses the three equations above, including no constant for Z-double-prime. “Safe zone” is the conventional label for the higher-score region, not a guarantee of safety. [1] [4] [5]
| Model | Distress zone | Gray zone | Higher-score or “safe” zone |
|---|---|---|---|
| Original Z | Below 1.81 | 1.81 to 2.99 | Above 2.99 |
| Z′ | Below 1.23 | 1.23 to 2.90 | Above 2.90 |
| Z″, no constant | Below 1.10 | 1.10 to 2.60 | Above 2.60 |
Here, exact boundary values are included in the gray zone as a conservative reporting convention. Retain unrounded calculations near a cutoff. A movement from 2.98 to 3.00 in original Z changes a label, but it does not demonstrate a sudden economic transformation.
You may also encounter 2.675 in discussions of original Z. Altman’s original study selected it as a single classification cutoff that minimized sample misclassifications; it is different from the two boundaries defining the gray interval. State which classification rule you use rather than merging them. [1]
Turning a score into a probability requires additional calibration for a specified population, default definition and horizon. Historical classification results do not establish a current company’s one-year or two-year default probability. The score also does not estimate creditor recovery after default.
Worked Example: From Financial Statements to Z
Consider ForgeWorks Manufacturing, a fictional publicly traded U.S. manufacturer. Use one full year’s sales and EBIT with its year-end balance sheet. All statement amounts are in USD millions. It has no preferred stock, noncontrolling interests or intangible assets. For this illustration, the share price and shares outstanding are measured on that same balance-sheet date.
| Item | Amount | Source or reconciliation |
|---|---|---|
| Current assets | 240 | Year-end balance sheet |
| Other assets | 360 | Year-end balance sheet |
| Total assets | 600 | 240 + 360 |
| Current liabilities | 160 | Year-end balance sheet |
| Long-term liabilities | 200 | Year-end balance sheet |
| Total liabilities | 360 | 160 + 200 |
| Book equity | 240 | 600 − 360 |
| Retained earnings | 120 | Part of book equity; other contributed equity is 120 |
| Annual EBIT | 48 | Annual income statement |
| Annual net sales | 720 | Annual income statement |
| Market equity | 360 | 30 million common shares × $12 per share |
The balance sheet reconciles: 600 = 360 + 240. Working capital is 240 − 160 = 80. Market equity is 360, while book equity is 240; those are different measurements of the same shareholders’ interest.
1. Calculate the Five Original-Model Ratios
| Input | Calculation | Weighted contribution |
|---|---|---|
| X1 | 80 / 600 = 0.133333… | 1.2 × (80 / 600) = 0.160 |
| X2 | 120 / 600 = 0.20 | 1.4 × 0.20 = 0.280 |
| X3 | 48 / 600 = 0.08 | 3.3 × 0.08 = 0.264 |
| X4,M | 360 / 360 = 1.00 | 0.6 × 1.00 = 0.600 |
| X5 | 720 / 600 = 1.20 | 1.0 × 1.20 = 1.200 |
| Original Z | Sum of weighted contributions | 2.504 |
2. Interpret the Score With the Original Zones
Z = 0.160 + 0.280 + 0.264 + 0.600 + 1.200 = 2.504. ForgeWorks is in the original model’s gray zone between 1.81 and 2.99. The appropriate response is to investigate its funding and operating resilience, not attach an invented probability to 2.504.
Review the contribution pattern as well as the total. Sales/assets contributes 1.200, while EBIT/assets contributes 0.264. Those are score contributions, not percentages of bankruptcy risk or a causal ranking of what management should change.
3. See Why the Variants Are Different Scales
Now consider two separate hypothetical companies: a private manufacturer with identical accounting figures, and a nonfinancial service company with the same working-capital, retained-earnings, EBIT and book-equity ratios. These are counterpart calculations for teaching the variants; ForgeWorks remains a public manufacturer.
Book-equity ratio: X4,B = 240 / 360 = 0.666666…
Z′ = 0.717 × (80/600) + 0.847 × 0.20 + 3.107 × 0.08 + 0.420 × (240/360) + 0.998 × 1.20.
Z′ = 0.0956 + 0.1694 + 0.24856 + 0.2800 + 1.1976 = 1.99116. This is gray under Z-prime’s 1.23 to 2.90 boundaries.
Z″ = 6.56 × (80/600) + 3.26 × 0.20 + 6.72 × 0.08 + 1.05 × (240/360).
Z″ = 0.874666… + 0.652 + 0.5376 + 0.700 = 2.764266…. This is above the no-constant model’s 2.60 higher-zone boundary.
The different labels illustrate why scores from different equations are not directly interchangeable. They do not establish that changing a company’s industry label improves its financial condition, or that ForgeWorks should use the service-company result.
4. Isolate the Market-Equity Effect
Return to ForgeWorks and hold every financial statement input and its 30 million shares fixed. Suppose its share price falls from $12 to $6. Market equity becomes 30 million × $6 = $180 million, so X4,M falls from 1.00 to 180 / 360 = 0.50.
The original score falls by 0.6 × (1.00 − 0.50) = 0.300, from 2.504 to 2.204. Both values are gray. The book-equity variants would not change directly from that price movement alone because their accounting inputs were held fixed. This sensitivity demonstrates the model’s market input; it does not translate a 50% stock-price decline into a bankruptcy probability.
Altman Z-Score vs Other Credit Tools
Use the score alongside tools that answer different questions. Its convenience comes from compressing information; the compression leaves out details that can determine whether a particular debt payment is made.
| Tool | Main question | What to keep separate |
|---|---|---|
| Altman score | Does the financial profile resemble historical distressed or nondistressed groups? | A classification score needs additional calibration to become a PD estimate. |
| Debt-to-equity and interest coverage | How much financing risk exists, and how well do earnings cover financing costs? | Individual diagnostics are not the full composite model. |
| Merton structural credit model | How do asset value, volatility and debt obligations relate under an option-based framework? | Its distance-to-default and probability conventions are not Altman zones. |
| Probability of default and expected loss | What is the chance of a defined default over a stated horizon, and the loss if it occurs? | Neither PD nor recovery is supplied directly by a raw Z-score. |
A company-level score also cannot determine whether one secured bond is more attractive than another unsecured bond. Maturity, collateral, guarantees, covenants, legal entity and price remain important. Those belong in the broader corporate bond credit analysis process.
How to Investigate After Calculating the Score
- Keep an input audit trail. Record the filing, reporting period, currency, consolidation scope, formula, equity definition, asset adjustments and market-value date. Reconcile EBIT if it is not reported directly; adjusted EBITDA is not an automatic substitute.
- Explain the drivers and trend. Recalculate prior periods consistently. Determine whether a movement comes from operating losses, working-capital changes, accumulated deficits, liability growth or a stock-price change.
- Examine cash availability. Read operating cash flow, restricted-cash disclosures and expected cash uses. A positive EBIT contribution does not establish that cash is available for the next payment.
- Map obligations over time. Compare maturities and interest payments with cash resources, committed facilities and realistic refinancing assumptions. Check covenant headroom, guarantees, leases, pensions and contingencies in the notes.
- Challenge the accounting and business assumptions. Review unusual transactions, revenue recognition, receivable quality, inventory, acquisitions and auditor commentary. Consider industry conditions and plausible adverse scenarios.
The result should be a clearer explanation of the company’s vulnerabilities and available responses. A high score should not stop the investigation when a near-term maturity, accounting concern or major event contradicts it. Altman’s own discussion treats scoring models as an additional analytical tool. [2]
Common Mistakes
- Substituting book equity into original Z. Use the properly reestimated book-equity model for the intended population.
- Using borrowings instead of total liabilities. Omitting nonborrowing liabilities changes the fourth ratio and can overstate the score.
- Entering 10 instead of 0.10 for a 10% ratio. The decimal-ratio equations above would exaggerate that contribution one hundredfold.
- Mixing annual and quarterly flows. One quarter’s sales are not directly comparable with a full year’s sales in an otherwise identical calculation.
- Mixing formulas and cutoffs. The original, private and no-constant nonmanufacturer zones are different; the emerging-market constant creates another convention.
- Discarding negative figures or missing data. Preserve genuine losses and deficits. An unavailable input is not zero.
- Treating a zone as a trading instruction. Distress screening does not establish valuation, recovery or the suitability of buying, selling or shorting a security.
Limitations of the Altman Z-Score
Historical Samples and Business Models
The original sample contained 66 manufacturers: 33 bankrupt firms and 33 comparison firms, with bankruptcies drawn from 1946 through 1965. That deliberately balanced research sample was not the ordinary economy’s bankruptcy rate. Its classification performance therefore cannot be read as the probability that a randomly selected company with a particular score will fail today. [1]
Accounting standards, financing structures, firm sizes and industry economics have changed. Later research has examined broader nonfinancial populations, including international samples, but model and country calibration still matter. No historical accuracy percentage establishes identical performance for a different industry, country or forecast horizon. [6]
Startups may have low retained earnings because they have had little time to accumulate profits. Asset-light companies, acquisition-heavy companies and businesses with unusual working-capital structures also require context. For banks and insurers, use dedicated capital, asset-quality, liquidity and sector analysis rather than treating Z-double-prime’s broader name as validation.
Accounting Quality and Missing Obligations
The formulas inherit errors in their inputs. In its June 2002 WorldCom action, the SEC alleged that the company had capitalized approximately $3.8 billion of costs instead of recognizing them as expenses, falsely portraying profitability. A calculation based on such reported figures cannot independently repair the underlying accounts. [7]
Enron supplies a related warning about missing obligations. Altman’s later case discussion explains that models using the publicly available accounts faced understated liabilities. The practical lesson is to investigate disclosures and economic obligations, even when a numerical output appears precise. These historical cases are illustrations of data limitations, not a claim that any particular score reliably detects fraud. [2]
Denominators, Timing and Unmodeled Events
Most inputs divide by assets, and the equity ratio divides by liabilities. Zero denominators make the corresponding ratios undefined; unusually small denominators can make them unstable. Asset impairments, acquisitions and disposals can change several ratios simultaneously, so a higher asset-turnover ratio after a write-down is not automatically evidence of better operations.
Negative book equity is in the numerator of X4,B. With positive liabilities, it creates a negative contribution rather than a division-by-negative-equity problem. Investigate its causes; accumulated losses, distributions and accounting adjustments can have different implications.
The score does not explicitly schedule debt maturities or model a sudden loss of financing, litigation, fraud revelation or operating disruption. Old statements and a recent market price can also describe different information sets. Use dated inputs and forward-looking cash analysis alongside the score.
Frequently Asked Questions
Sources
- Edward I. Altman, Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy, Journal of Finance, September 1968. Original sample, method and historical original-Z boundaries; original-paper text read through a third-party transcription.
- Edward I. Altman, Corporate Distress Prediction Models in a Turbulent Economic and BASEL II Environment, NYU Stern, September 2002. Author-presented decimal formula, private and nonmanufacturer adaptations, calibration and historical case limitations.
- U.S. Securities and Exchange Commission, Beginners’ Guide to Financial Statements, February 2007. Statement categories, stock-versus-flow timing, cash flows and notes.
- Edward I. Altman, Predicting Financial Distress of Companies: Revisiting the Z-Score and ZETA Models, July 2000, original-paper transcription mirror. Private-company equation and historical 1.23 to 2.90 gray interval. Historical NYU PDF was inaccessible during research.
- Justin Lord and coauthors, Predicting Nursing Home Financial Distress Using the Altman Z-Score, Inquiry, July 2020. Published corroboration of the historical no-constant four-ratio formula and 1.10/2.60 boundaries; its separate nursing-home reestimation is not used here.
- Edward I. Altman and coauthors, Financial Distress Prediction in an International Context, 2017. Public abstract on broader nonfinancial populations and country-specific estimation; full text was not accessed.
- U.S. Securities and Exchange Commission, WorldCom, Inc., Litigation Release 17588, June 27, 2002. Contemporaneous allegations concerning capitalization of expenses and overstated profitability.
Sources accessed November 15, 2026. All ForgeWorks and counterpart figures are independently calculated hypothetical examples, not actual company results or current forecasts.
Disclaimer
This article is for educational and informational purposes only and does not constitute investment, credit or legal advice. Historical screening models can misclassify companies and do not determine current default probabilities, recoveries or investment suitability. Review original financial statements and consult a qualified professional about your circumstances.
