Z-Score Default Prediction: From Cutoff Zones to GM Bankruptcy

Corporate Financial Distress, Restructuring, and Bankruptcy by Edward I. Altman, Edith Hotchkiss, and Wei Wang (Wiley, ISBN 978-1-119-48180-5)

Previous: Altman Z-Score at 50 | Next: Distress Prediction for External Analysts

The first half of Chapter 10 traced the history of credit scoring. This second half is where Altman gets practical: how to turn a Z-Score into a probability of default, why the old cutoff zones do not work anymore, and what happened when he took the model to Congress during the auto bailout debate.

Safe, Gray, and Distress Zones

The original model used three zones:

• Z > 2.99: Safe zone
• 1.81 < Z < 2.99: Gray zone
• Z < 1.81: Distress zone

These cutoffs came from a sample of 66 firms in the 1960s. Holdout tests from 1969 through 1999 still showed 85%+ accuracy one year before bankruptcy. Impressive. But Altman admits it is “unfortunate” that the zones never got updated. Credit markets changed radically. The classification of “bankrupt or not” is no longer enough for most applications.

A firm at 1.81 and a firm at 1.79 are nearly identical, but one sits in the gray zone and one in distress. The real goal is probability of default (PD) and timing, not a binary label.

Credit Quality Has Shifted Over 50 Years

Median Z-Scores by S&P rating tell the story. AAA/AA median scores fell from 5.20 (1996-2001) to 4.30 (2017). Only two AAA-rated firms remained in 2017: Microsoft and Johnson & Johnson. Single-B median Z-Scores dropped from 1.87 to 1.65. Remember, 1.81 was the distress cutoff in 1966. Today, a huge chunk of the high-yield market issues at single-B, and only about 28% of B-rated bonds default within five years.

The Type II error (predicting default when the firm survives) may have risen from 5% originally to 25-30% recently. Altman’s recommendation: stop using 1.81 as a hard cutoff. Use bond-rating equivalents (BREs) mapped from current median Z-Scores by rating category instead.

Two Methods for Estimating PD

Method 1: Score the firm, map to a BRE, then apply mortality rates or cumulative default rates by rating. Altman’s mortality approach (1989) tracks bonds from original issuance rating, weighted by dollars. A BB-rated issue has first-year, second-year, and third-year marginal default rates of 0.92%, 2.04%, and 3.85%. Cumulative mortality for B-rated bonds reaches about 36.6% over ten years.

Method 2: Run a logistic regression directly on firm data to get a PD between 0 and 1, then map back to a rating equivalent.

Altman prefers Method 1 for newly issued debt (mortality rates) and cumulative default rates for existing portfolios. Logistic models give direct PDs but depend heavily on the sample used to build them.

Z’-Score and Z"-Score: When to Use Which

The Z’-Score swaps book equity for market equity, built for private firms. The Z"-Score drops sales/assets, adds a constant (3.25), and works across manufacturing and non-manufacturing, developed and emerging markets. It handles retailers and service firms better than the original.

The Sears example from 2016 shows why this matters. The original Z-Score gave Sears a B- equivalent (1.3). The Z"-Score put it at D/CCC. Sears was a retailer with high sales relative to assets, which inflated the original score. Z"-Score caught the distress the manufacturing-focused model missed.

GM: Testimony That Congress Ignored

In December 2008, Altman testified before the House Finance Committee on whether to bail out GM and Chrysler or push them into Chapter 11. His Z-Score data told a clear story. GM sat in CCC territory for years while still rated investment grade in 2005. By December 2008, the score was -0.63, deep in default territory.

He recommended Chapter 11 with a $50 billion DIP loan from the government. The House voted to continue the bailout. The Senate voted no. President Bush provided interim funds anyway. GM filed Chapter 11 on June 1, 2009, with exactly the $50 billion DIP loan Altman had proposed six months earlier. GM emerged in 43 days and eventually returned to investment-grade ratings.

But here is the twist: by 2014, agencies rated GM BBB while the Z-Score still showed single-B. The model and the agencies disagreed again.

2007 vs. 2016: Are We Safer Now?

Comparing high-yield issuers across periods, average Z-Scores were nearly identical: 1.95 (B+) in 2007 and 1.97 in 2016. Z"-Scores were actually higher in 2007. Statistical tests found no significant difference. The average risky borrower in 2016 looked about as healthy as one in 2007, right before the financial crisis. Whether that is reassuring or terrifying depends on your priors.

Energy Sector: The Model Still Works

During the 2015-17 energy bust, more than half of all defaults came from energy and mining. Testing Z- and Z"-Scores on 31 bankrupt firms: 84% had D-rated equivalents one or two quarters before filing. The original manufacturing model, applied to oil and gas companies it was never designed for, still caught the distress.

My Take

This is the chapter where Altman stops being a professor and starts being a witness. The GM testimony story is the highlight. He told Congress exactly what would happen. They did the opposite. Then events proved him right, down to the dollar amount of the DIP loan.

The deeper lesson is about model maintenance. A 50-year-old cutoff score is a historical artifact, not a live risk tool. The BRE mapping and mortality rate approach are how you keep the Z-Score relevant. And the Sears vs. GM examples show that picking the right variant (Z vs. Z" vs. Z’) matters as much as running the calculation at all.

Previous: Altman Z-Score at 50 | Next: Distress Prediction for External Analysts