Altman Z-Score: Reading Bankruptcy Risk From a Balance Sheet

Most of us read a results release from the top: revenue, profit, margin. The Altman Z-score reads it from the bottom up. It takes five plain ratios from the balance sheet and profit and loss account, weights them, and spits out one number that tries to answer a blunt question: how close is this company to going bust?

It’s an old tool, published in 1968, and it has real flaws. But it is quick, it uses numbers every listed company already discloses, and it forces you to look at working capital, accumulated profits and debt at the same time. That alone makes it worth knowing.

Where the Altman Z-score comes from

Edward Altman, then an assistant professor of finance at New York University, published the formula in 1968 in the Journal of Finance. According to the Wikipedia entry on the Altman Z-score, his starting sample was 66 publicly held manufacturers. Half had filed for bankruptcy. The other half had survived and were matched to them by industry and approximate size. Firms with assets under $1 million were left out.

He then used a statistical method called discriminant analysis. In simple terms, it searches for the combination of ratios that best separates the two groups. Earlier researchers, notably William Beaver, had tested ratios one at a time. Altman’s step forward was to weigh several of them together. The result was a weighted sum, which is why the score moves when any one of the five inputs moves.

How far apart did the groups end up? Wikipedia notes that the bankrupt firms averaged a score of about −0.25, while the survivors averaged about +4.48. That is a wide gap, and it explains why the tool caught on.

The formula, one ratio at a time

The original model, as given by Wikipedia and by Investopedia’s Altman Z-score explainer, is:

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

Here is what each piece measures, and what we think it is really asking.

• X1 = working capital ÷ total assets. Working capital is current assets minus current liabilities. The question here is simple: can the company pay what falls due in the next year from what it will turn into cash in the next year? A negative X1 drags the whole score down.
• X2 = retained earnings ÷ total assets. This is profit the business has earned and kept over its life. A young company, or one that has burned through past profits, scores low. Use the retained-earnings figure itself, not every reserve on the balance sheet, because money raised from shareholders is not earned profit.
• X3 = earnings before interest and taxes (EBIT) ÷ total assets. Operating earning power, before the effects of debt and tax. Notice it carries the heaviest weight, 3.3. A company that cannot earn on its assets runs out of road fast.
• X4 = market value of equity ÷ total liabilities. Market capitalisation against everything the company owes. This is the only input that comes from the stock market rather than the books, so a falling share price lowers the score even before the accounts look bad.
• X5 = sales ÷ total assets. Asset turnover. It tells you how hard the assets are working. Wikipedia points out that it varies a lot from industry to industry, which is one reason the score travels badly across sectors.

The Z-score doesn’t care how good the story is. It cares whether the bills can be paid, whether profits were ever kept, and whether the market still values the equity above the debt.

Reading the zones

For the original model, Wikipedia lists three zones:

• Above 2.99: the “safe” zone.
• Between 1.81 and 2.99: the “grey” zone, where the score doesn’t give a clear verdict.
• Below 1.81: the “distress” zone.

Investopedia rounds these to 1.8 and 3. It also adds an important update. In a 2019 lecture titled “50 Years of the Altman Score”, Altman said recent data suggests 0, not 1.8, is the level at which investors should really worry. So treat 1.81 as an early warning and a score near or below zero as a loud one.

A worked example in rupee crore

Let’s take a hypothetical mid-sized auto-parts maker. It’s a manufacturer, so the original formula fits. All figures below are made up for illustration and are in ₹ crore.

Year 1: a healthy balance sheet

• Total assets: 2,000
• Current assets: 900; current liabilities: 600, so working capital is 300
• Retained earnings: 500
• EBIT: 240
• Market capitalisation: 1,800; total liabilities: 1,200
• Sales: 2,600

Now the ratios:

• X1 = 300 ÷ 2,000 = 0.15, weighted 1.2 × 0.15 = 0.180
• X2 = 500 ÷ 2,000 = 0.25, weighted 1.4 × 0.25 = 0.350
• X3 = 240 ÷ 2,000 = 0.12, weighted 3.3 × 0.12 = 0.396
• X4 = 1,800 ÷ 1,200 = 1.50, weighted 0.6 × 1.50 = 0.900
• X5 = 2,600 ÷ 2,000 = 1.30, weighted 1.0 × 1.30 = 1.300

Add them up and Z comes to about 3.13. That sits just above 2.99, in the safe zone, though not by a huge margin.

Year 2: the same company after a rough patch

Say demand slows, a big customer delays payments, and the company funds the gap with short-term borrowing. The share price halves and then some.

• Total assets: 2,200
• Current assets: 850; current liabilities: 900, so working capital is −50
• Retained earnings: 450 (a loss after interest ate into past profits)
• EBIT: 66
• Market capitalisation: 700; total liabilities: 1,550
• Sales: 2,400

• X1 = −50 ÷ 2,200 = about −0.023, weighted about −0.027
• X2 = 450 ÷ 2,200 = about 0.205, weighted about 0.286
• X3 = 66 ÷ 2,200 = 0.03, weighted 0.099
• X4 = 700 ÷ 1,550 = about 0.452, weighted about 0.271
• X5 = 2,400 ÷ 2,200 = about 1.091, weighted about 1.091

Z drops to about 1.72, below 1.81 and into the distress zone.

Look at where the damage came from. The market-value term fell from 0.900 to about 0.271. The EBIT term fell from 0.396 to 0.099. Those two moves alone take away roughly 0.93 points. Working capital turning negative did the rest. Sales barely changed the picture. That’s the useful part of the exercise: the total tells you there’s a problem, and the five pieces tell you where to dig.

A falling Z-score is a reason to open the notes to accounts, not a reason to panic. The five parts usually point straight at the page you need to read.

Versions for non-manufacturers and emerging markets

The original was built on manufacturers, and the sales-to-assets term punishes asset-heavy businesses that aren’t factories. Altman later produced other versions. Wikipedia gives the four-ratio model for non-manufacturers as:

Z” = 6.56 X1 + 3.26 X2 + 6.72 X3 + 1.05 X4

Two things change. The sales term is gone, and X4 uses the book value of equity rather than market value. For emerging markets, the same equation adds a constant of 3.25. The zones listed alongside these versions are above 2.6 for safe, 1.1 to 2.6 for grey, and below 1.1 for distress.

Run our Year 1 company through it, with book equity of 800 (total assets of 2,000 less liabilities of 1,200). X4 becomes 800 ÷ 1,200, about 0.667. The weighted terms are 0.984, 0.815, 0.806 and 0.700, which add to about 3.31. Still safe, but you can see how much the answer depends on which version you pick. Always say which formula you used when you compare scores.

How Indian defaulters scored

Wikipedia reproduces a table from a 2014 Symbiosis International University study by Swati Khatkale on Indian credit rating agencies, covering large defaults between April 2005 and March 2013. A few numbers stand out:

• Kingfisher Airlines had a Z-score of −2.36 in the year of its December 2011 default, and 1.85 a year earlier.
• Suzlon Energy scored −0.11 in the year of its June 2012 default, and 1.23 a year before that.

The study’s conclusion, as quoted there, is that the Z-score flagged defaults at Royal Orchid Hotel, Arvind Products and Suzlon that the credit ratings at the time didn’t reflect. It’s one study, and a handful of cases, so we wouldn’t lean on it too hard. But it shows the score isn’t just an American curiosity.

How accurate is the Altman Z-score?

The headline numbers are decent. Wikipedia says the first test found the model 72% accurate in predicting bankruptcy two years ahead. Later tests over three periods up to 1999 put accuracy at roughly 80% to 90% one year before the event, with false alarms (firms flagged as bankrupt that didn’t fail) of about 15% to 20%. Investopedia, describing Altman’s samples of distressed firms from 1969 to 1999, quotes an accuracy range of 82% to 94%.

There are serious caveats, though.

• It describes more than it predicts. Critics argued early on that showing failed and healthy firms have different ratios is not the same as proving the ratios predict failure.
• Not for banks and NBFCs. Wikipedia is clear that neither the Altman models nor other balance-sheet models are recommended for financial companies, because their balance sheets are opaque and they use a lot of off-balance-sheet items. Lending is their business, so their liabilities look huge by design.
• Sector matters. Asset turnover differs so much between, say, a retailer and a power utility that the original formula can mislead across industries.
• Newer models lean on market data. Modern academic default models rely heavily on market-based data rather than accounting ratios alone.

Investopedia also notes a market-wide use. Altman calculated that the median Z-score of the companies he studied in 2007 was 1.81, which he equated to a B rating, and he expected a wave of corporate defaults. The crisis that followed actually started in mortgage-backed securities, but corporate defaults in 2009 then ran at the second-highest rate in history.

Using it on Indian results, practically

Here’s how we’d put it to work without overthinking it.

1. Use full-year numbers. EBIT and sales need a whole year to mean anything against total assets. Audited annual accounts are the cleanest source; trailing four quarters work if you add them up carefully.
2. Pick the right version. Original Z for manufacturers, Z” for most others, and skip financial companies altogether.
3. Track the trend, not one reading. A company sliding from 3.5 to 2.4 to 1.9 over three years is telling you more than a single 2.4 ever could.
4. Break it into parts. When the score drops, find which term moved. A share-price fall alone is a different story from a working-capital squeeze.
5. Pair it with other checks. The Piotroski F-score looks at whether fundamentals are improving, and free cash flow yield tells you whether profits are turning into cash. If you’re new to reading results, start with our guide on how to analyse quarterly results of a company.

What to take away

The Altman Z-score is a five-ratio stress test. It mixes liquidity, accumulated profit, operating earnings, market confidence and asset use into one figure, with 1.81 and 2.99 as the classic dividing lines for manufacturers. It’s quick, it’s transparent, and it has flagged real trouble in India and abroad. It’s also old, sector-sensitive and useless for banks. Use it as a smoke detector, and when it goes off, go and read the balance sheet line by line.

Disclaimer: This article is for educational purposes only and is not investment advice. The company and all figures in the worked example are hypothetical. Investing involves risk of loss; consult a SEBI-registered investment adviser before making investment decisions.