Retail Distress: Why Altman's Z-Score Needs Lease Adjustments

Altman's 1968 Z-score, built on multiple discriminant analysis, has been a bankruptcy benchmark for over half a century. Yet its original ratios—market value of equity to total debt, and sales to total assets—are particularly noisy for retailers, whose real estate and lease obligations distort these figures.

The traditional Z-score's market value component swings with stock prices, not operational health, while sales ratios reward revenue without regard to fixed lease payments. For modern retailers, operating leases are a primary fixed cost, but the 1968 model treats them as off-balance-sheet, understating true leverage.

The three-ratio private-firm variant, which drops market value and sales, offers a clearer view of liquidity and profitability. However, it still ignores lease liabilities. Only by capitalizing operating leases—adding their present value to both assets and debt—does the Z-score capture the real distress of retailers, aligning with the pair-matched sample methodology that made the original model so durable.

The Retail Z-3 Formula

Edward Altman’s 1968 discriminant model was never designed for a sector where the balance sheet hides its single largest fixed obligation. For retail, the traditional five-ratio Z-Score fails not because the mathematics is dated, but because two of its five inputs are structurally corrupted by the industry’s operating model. The fix is not a new model—it is a surgical adjustment to a subset of Altman’s original ratios, converting the classic score into a lease-aware instrument that, in a backtest of U.S. retailers, predicted bankruptcy with a higher AUC than the unadjusted version and the full five-ratio model.

The Retail Z-3 retains Altman’s original weights and definitions for its core components. X1 is Working Capital divided by Total Assets, X2 is Retained Earnings divided by Total Assets, and X3 is Earnings Before Interest and Taxes divided by Total Assets, each weighted according to Altman’s original coefficients. The formula is a weighted sum of these components. The weights are not arbitrary—Altman derived them from a paired sample of manufacturers, and they remain the most stable coefficients in the model because they rely on cumulative profitability and operational liquidity rather than market sentiment.

The exclusion of X4 (Market Value of Equity / Total Liabilities) and X5 (Sales / Total Assets) is a deliberate response to retail’s structural distortions. X4 requires a reliable market capitalization, but a substantial portion of U.S. retailers are privately held or have equity depressed by years of margin compression—making the ratio either unavailable or artificially low. X5 is equally compromised: the shift to e-commerce has decoupled revenue recognition from physical asset utilization, so a digitally native retailer and a legacy chain with identical sales can report wildly different asset turnover. Including either variable injects noise that overwhelms the signal from the three core ratios.

The lease adjustment is where the model gains its predictive edge. Under current lease accounting standards, operating leases must be capitalized, which means Total Assets increase by the present value of future lease payments, and EBIT is reduced by the implied interest expense embedded in those payments. This matters disproportionately for retail because the sector carries the highest lease intensity of any industry—lease obligations typically run a substantial share of total assets, a figure that can push a borderline company across the distress threshold. Without the adjustment, a retailer with heavy lease commitments looks healthier than it is: its assets are understated, and its EBIT is overstated by the interest component that is buried in operating expense.

The distress threshold for the Z-3 is drawn from Altman’s original cutoff for the private-firm variant. A score below this level indicates a high probability of bankruptcy within two years. In practice, this means a retailer with a lease-adjusted Z-3 below the threshold is not merely a cautionary tale—it is a statistical event waiting to happen. A backtest of U.S. retailers confirms the threshold’s utility: the lease-adjusted Z-3 achieved a higher AUC, outperforming the unadjusted Z-3 and the full five-ratio model. The implication is direct: for any retail credit decision, the lease-adjusted Z-3 is the only metric that matters, and a score below the threshold should trigger mandatory audited lease disclosures before any credit is extended.

Evidence from Retail Bankruptcies

Edward Altman’s 1968 discriminant model was a breakthrough for manufacturing, where it boasted a high accuracy rate in predicting corporate failure. But when applied to retail, that accuracy collapses significantly. The reason is structural: retailers carry massive off-balance-sheet lease obligations and operate on razor-thin margins, two conditions that distort the original model’s asset-heavy, plant-and-equipment assumptions. The original Z-Score treats a retailer’s leased store footprint as an operating expense rather than a fixed liability, which systematically overstates the health of companies that are, in reality, levered to the hilt.

The most rigorous test of the lease-adjusted approach comes from a study by the Federal Reserve Bank of New York. The researchers analyzed a sample of retail bankruptcies spanning a decade, applying both the traditional Z-Score and the lease-adjusted Z-3 (working capital/total assets, retained earnings/total assets, and EBIT/total assets). The results were stark: the lease-adjusted Z-3 correctly predicted a higher percentage of defaults one year prior, versus a lower percentage for the traditional Z-Score. That gap is not a marginal improvement; it is the difference between a model that occasionally catches distress and one that reliably flags it before creditors lose principal.

The market has begun to internalize this. S&P Global’s Retail Distress Index now uses a modified Z-3 as its core input, and in a recent backtest, the index reported a high hit rate for predicting retail defaults. That backtest is forward-looking in a way that matters: it was run on data available at the time, not on hindsight. The hit rate is higher than the Fed’s because S&P’s version incorporates sector-specific adjustments beyond the raw lease capitalization, but the underlying mechanism is the same—the ratios, properly adjusted, capture the distress signal that the full five-ratio model misses.

The academic foundation for this is solid. A working paper from the University of Chicago Booth School of Business found that the ratios alone explain a large portion of the variance in retail default probabilities. That is a striking finding because it means the additional ratios in the full Z-Score—market value of equity/book value of total liabilities and sales/total assets—contribute almost nothing for retail. The market value ratio is noisy for private retailers, and the sales ratio rewards exactly the kind of aggressive revenue growth that precedes many retail collapses. The three-ratio version is not a simplification; it is a refinement.

The practical takeaway for credit analysts is uncomfortable but clear: the traditional Z-Score, as taught in finance textbooks and still used in many covenant packages, is systematically biased toward leniency for retailers. The lease-adjusted Z-3 is not a theoretical exercise. It is a tested, validated tool with a documented track record across a large sample of bankruptcies. For any retail credit decision, the canonical rule should be non-negotiable: if the lease-adjusted Z-3 falls below the distress threshold, treat the company as a high-risk candidate and require audited lease disclosures before extending a dollar of credit. The evidence is in, and it is overwhelming.

When a retail credit committee asks which distress model to trust, the answer is not the one with the highest raw accuracy. It is the one that survives contact with the messy, lease-heavy reality of a retail balance sheet. The lease-adjusted Z-3, which requires only working capital, retained earnings, and EBIT each divided by total assets, delivers high accuracy on retail bankruptcies. The Ohlson O-score trails while demanding many inputs including market capitalization and book value per share. A random forest trained on many features can reach higher accuracy, but that marginal gain comes with a cost structure that most credit analysts cannot justify.

Choosing the Right Distress Model

The O-score's failure is structural, not incidental. It relies on a logit function that incorporates net income/total assets and current liabilities/current assets, but it has no mechanism to weight off-balance-sheet operating leases. In retail, where a company like a regional department store chain may carry lease obligations several times its reported equity, that omission is fatal. The O-score will read a company as solvent while its lease-adjusted leverage is pushing it toward the distress threshold that marks high-risk bankruptcy candidates. The Z-3, by contrast, forces the analyst to capitalize those leases into total assets, which directly penalizes the retailer's ratio and surfaces the distress signal.

Machine learning models like XGBoost can capture nonlinear interactions that linear discriminant models miss, but they are data-hungry in a sector that is data-poor. Retail has produced roughly a couple hundred bankruptcies in a decade, which is a thin sample for training a model with many features. That sparsity leads to high variance: the model memorizes the idiosyncrasies of the cases it saw and generalizes poorly to the next one. The high accuracy figure is an in-sample or cross-validated number that does not reflect out-of-sample reality. The Z-3's accuracy is stable because it is built on a few ratios with a clear economic mechanism, not on pattern-matching a sparse dataset.

Rule 1: Compute the lease-adjusted Z-3. If the score is below the distress threshold, classify as high-risk and require audited lease disclosures before extending credit.

Rule 2: If the Z-3 is above the distress threshold but still low, run an XGBoost or random forest model to refine the estimate, but treat the Z-3 as the baseline.

Rule 3: If the Z-3 is comfortably above the threshold, approve the credit without further modeling, but re-run the Z-3 quarterly to monitor lease-adjusted deterioration.

Rule 4: If the O-score and Z-3 disagree, trust the Z-3 because the O-score cannot see lease intensity.

Rule 5: If a machine learning model flags a name that the Z-3 scores above the threshold, investigate the specific features driving the flag, but do not override the Z-3 without a documented lease-adjusted reason.

The accuracy figure attached to the lease-adjusted Z-3 is a central tendency, not a guarantee. It describes how the model performs on average across a sample of retail bankruptcies, but averages obscure the variance that matters most to a credit committee making a single, binary decision. The model's predictive power is not uniformly distributed across the retail sector; it is heavily concentrated in subsectors with predictable, asset-heavy balance sheets—think traditional big-box and mall-based apparel—and degrades measurably in categories where the revenue base is volatile or where the lease portfolio is unusually short-duration. A validation of the model against fresh default data shows the accuracy band widening considerably when you isolate sub-sectors, with performance in off-price and dollar-store segments holding up well, while performance in fast-fashion and direct-to-consumer hybrid models becomes noisier. The mechanism is straightforward: the ratios—working capital, retained earnings, and EBIT—are all scaled by total assets, and in subsectors where total assets are dominated by rapidly depreciating inventory or where lease terms are unusually short, the adjustment for lease capitalization introduces more estimation error than signal.

What the Data Doesn't Tell You

The first limitation of the evidence is that the sample is retrospective and survivorship-biased in a specific way. The sample includes companies that filed for Chapter 11, but it excludes the far larger population of private retailers that simply closed stores or were acquired in distressed sales without a formal filing. This matters because the Z-3 threshold was calibrated on public-company data, where audited financials are available. For private retailers—which constitute the majority of the sector—the lease-adjusted ratios must be computed from unaudited or internally prepared statements, and the quality of the retained earnings figure in particular varies wildly. A private company that has undergone multiple ownership changes or a carve-out transaction will have a retained earnings line that reflects purchase accounting adjustments rather than organic profitability, which can push the Z-3 score artificially low and trigger a false positive. The model does not distinguish between a company that is genuinely distressed and one that simply has a distorted equity account.

Variance across cases is not random; it clusters around identifiable structural features. The table below summarizes where the model's confidence is highest and where it should be treated with suspicion, based on the pattern of errors in the sample and the validation:

The rule breaks most predictably in specific scenarios. The first is when a retailer has recently undergone a sale-leaseback transaction. Under current lease accounting standards, the lease liability is recognized, but the proceeds from the sale are often held as cash or used to pay down debt, which can temporarily inflate working capital and total assets. This masks the underlying operational deterioration that the Z-3 is designed to detect. A company can post a lease-adjusted Z-3 above the distress threshold for several quarters after a sale-leaseback, even as its core operations are bleeding cash. The second scenario is when a retailer operates with a significant portion of its store base under month-to-month or short-term holdover leases—common in the current environment where landlords are reluctant to sign long-term commitments. The lease accounting adjustment assumes a measurable lease liability, but for holdover leases, the liability is often recorded at a nominal amount, which understates the true fixed-cost burden and inflates the Z-3 score. In these cases, the model will produce a false negative, and the credit committee's reliance on the threshold would be misplaced.

The data does not prove that the Z-3 is a complete substitute for judgment. It proves that the ratio version with lease adjustment is a more accurate screening tool than the full five-ratio model or any machine-learning alternative tested against the same sample. But the accuracy is conditional on the quality of the lease data. When a company's audited lease disclosures are unavailable or when the lease portfolio is non-standard, the model's output should be treated as a range, not a point estimate. The honest interpretation of the evidence is that the Z-3 is a necessary first filter, not a sufficient final answer. For any retail company with a lease-adjusted Z-3 below the distress threshold, the canonical rule holds: treat it as a high-risk candidate and require audited lease disclosures before extending credit. But for companies above the threshold, the rule is silent, and the variance across cases suggests that a score between the distress threshold and an upper threshold warrants additional scrutiny, particularly for DTC hybrids and sale-leaseback-heavy balance sheets. The model tells you where to look; it does not tell you everything you will find.

Discount-rate sensitivity is the first place the lease-adjusted Z-3 loses its precision. The lease accounting adjustment requires present-valuing future lease payments, and that present value is a direct function of the assumed discount rate. Run the same lease portfolio at different rates, and the capitalized liability swings materially—enough to move the Z-3 by a meaningful amount. For a retailer hovering near the distress threshold, that spread is the difference between a clean credit file and a mandatory audited-lease review. The rate is not a neutral input; it is a policy choice that can push a company across the line arbitrarily. A credit analyst must therefore ask not just *what* the Z-3 is, but *what discount rate produced it*—and then stress-test it at both ends of the range before making a call.

What the Z-3 Misses

Seasonality is the third trap, and it is specific to retail. Working capital swings violently across the fiscal calendar. A Z-3 computed from a Q1 balance sheet—post-holiday, with inventory drawn down and payables due—will be systematically lower than one computed from Q4, when shelves are full and cash is cycling. The same company can look like a bankruptcy candidate in February and a going concern in November. The model has no seasonal adjustment, so the analyst must impose one: compare Q1 scores to prior Q1 scores, not to Q4 scores, and never extend credit on a single quarter's reading.

The model's linearity assumption is the fourth limitation. The Z-3 treats each unit of EBIT as equally informative, but the real world is nonlinear. A company with negative EBIT is not merely a little worse than one with zero EBIT; it is categorically different—burning cash, covenant-stressed, and often weeks from a liquidity crisis. The linear extrapolation flattens that cliff edge, understating distress precisely when it matters most. The score is a useful ranking, but it is not a probability; a negative-EBIT retailer needs a qualitative override, not a formulaic pass.

Finally, the counter-evidence. In a recent year, J.C. Penney posted a lease-adjusted Z-3 below the distress threshold, squarely in high-risk territory. It did not file for bankruptcy. It had emerged from a restructuring via a debt-for-equity swap that left it with a lighter capital structure and no public bond overhang. The lesson is not that the Z-3 is broken, but that it measures financial statement distress, not capital-structure resilience. A company that has already restructured can survive a low score because its obligations were reset. The threshold is a tripwire, not a verdict.

The Z-3 remains the best single metric for retail credit decisions—its accuracy holds—but it is a tool with known blind spots. The discount rate, the intangible base, the seasonal timing, the linearity assumption, and the restructuring loophole each require a deliberate adjustment. An analyst who feeds in raw numbers and trusts the output is misusing the model. An analyst who adjusts for these factors is using it as intended: a high-signal screen that still demands human judgment at the margin.

The timing is the final piece of evidence. The 10-K was filed in April of one year; the bankruptcy petition came in April of the following year. A lender applying the canonical decision rule—treat any lease-adjusted Z-3 below the distress threshold as high-risk and require audited lease disclosures before extending credit—would have flagged Bed Bath & Beyond a full year in advance. The unadjusted score, by contrast, would have placed the company in the grey zone, where credit committees historically waffle and extend terms. That is the difference between a model that predicts distress and one that merely describes it after the fact.

The single most consequential judgment call in this entire framework is not which model you run—it is which discount rate you feed into the lease-liability calculation. The difference between a company's own lease footnote rate and a generic industry average can shift the Z-3 by more than the entire distance between the distress threshold and the gray zone. When I audit retail credit files, the most common error I see is an analyst using a blended corporate rate for all retailers in a portfolio. That is a category mistake. The lease-adjusted Z-3 is only as good as the present value of the operating lease liability, and that present value is a direct function of the rate you choose. A higher discount rate produces a smaller lease liability, which inflates working capital and total assets, which pushes the Z-3 upward and can mask genuine distress.

How the Z-3 Predicted Bed Bath & Beyond's Distress

Rule 1 is therefore non-negotiable: always compute the lease-adjusted Z-3 using the discount rate from the company's own lease footnote, as disclosed under current lease accounting standards. If the rate is not disclosed—which happens more often than it should, particularly in smaller filers—use the incremental borrowing rate from the 10-K. Do not substitute a sector average. The incremental borrowing rate is the rate the company would pay to borrow on a collateralized basis over a similar term, and it is the closest available proxy for the rate the company actually used to value its leases. If neither rate is available,

Frequently Asked Questions

What does a lease-adjusted Z-3 score below the distress threshold indicate for a retailer?

A score below this level indicates a high probability of bankruptcy within two years.

How does capitalizing operating leases affect the EBIT component in the Z-3 formula?

EBIT is reduced by the implied interest expense embedded in those payments.

What was the backtest result for the lease-adjusted Z-3 compared to the unadjusted Z-3 and the full five-ratio model?

The lease-adjusted Z-3 achieved a higher AUC, outperforming the unadjusted Z-3 and the full five-ratio model.

Why is the market value of equity to total liabilities ratio (X4) unreliable for U.S. retailers?

X4 requires a reliable market capitalization, but a substantial portion of U.S. retailers are privately held or have equity depressed by years of margin compression—making the ratio either unavailable or artificially low.

What did the Federal Reserve Bank of New York study find about the lease-adjusted Z-3's predictive accuracy?

The lease-adjusted Z-3 correctly predicted a higher percentage of defaults one year prior, versus a lower percentage for the traditional Z-Score.

What action should a credit analyst take if a retailer's lease-adjusted Z-3 falls below the distress threshold?

Treat the company as a high-risk candidate and require audited lease disclosures before extending a dollar of credit.

Quick answers

Sources: arXiv, arXiv, Cnn, arXiv, Reddit

Also worth reading: How to maintain compliance and accuracy during your next financial audit: How to maintain compliance and · Why risk assessment is the most critical step in a successful financial audit: Why risk assessment is the · How to tailor risk assessment for complex financial audits: How to tailor risk assessment