Default Prediction

How Default Prediction Works: The Components of Credit Risk

Professional default prediction is rarely a single number; it is a calculation that factors into a broader "Risk Equation." To understand how a bank views a borrower, you must understand the three primary components of credit risk: 1. Probability of Default (PD): This is the "Prediction" itself. It is the percentage chance that the borrower will miss a payment or file for bankruptcy within a certain timeframe. 2. Loss Given Default (LGD): This measures the "Severity" of the loss. If the borrower defaults, how much of the money will be lost? For a secured loan (like a mortgage), the LGD is low because the bank can sell the house. For an unsecured loan (like a credit card), the LGD is often near 100%. 3. Exposure at Default (EAD): This is the "Amount" at risk. For a term loan, this is the current balance. For a credit line, it includes the amount already borrowed plus the amount the customer is likely to draw down as they start to enter financial distress. By multiplying these three factors (PD * LGD * EAD), a financial institution arrives at the "Expected Loss." Default prediction models focus heavily on the PD component. They look for "Liquidity Ratios" (can the borrower pay today?), "Solvency Ratios" (can they pay in the long run?), and "Profitability Trends." In the modern era, these models also include "Alternative Data" such as a CEO's sentiment in earnings calls or even satellite imagery of a retailer's parking lot to see if customer traffic is declining before it shows up in the quarterly reports.

Comparison: Structural vs. Reduced-Form Models

There are two primary mathematical philosophies used to predict financial failure in the corporate world.

Foundational Model: The Altman Z-Score

No discussion of default prediction is complete without the "Altman Z-Score." Developed in 1968, this model proved that you could predict bankruptcy with over 70% accuracy using just five simple accounting ratios. These ratios measure working capital, retained earnings, earnings before interest and taxes (EBIT), market value of equity, and sales. By assigning a "Weight" to each ratio, the formula produces a single number that places a company in one of three zones: the "Safe Zone," the "Grey Zone," or the "Distress Zone." While the original Z-Score was designed for manufacturing companies, it has been adapted for service firms and emerging markets. Its enduring popularity is due to its "Transparency"—unlike many modern "Black Box" AI models, an analyst can see exactly which financial weakness is driving the distress. If a company's Z-Score is falling because its sales-to-assets ratio is declining, it suggests an "Efficiency Problem." If it is falling because of low retained earnings, it suggests a "Sustainability Problem." This diagnostic power makes the Z-Score a foundational tool for value investors looking to avoid "Value Traps" and for credit analysts performing their initial due diligence.

Important Considerations: The Limitations of Prediction

Despite the mathematical sophistication of modern models, default prediction is not a "Crystal Ball." The most significant limitation is "Model Risk"—the danger that the assumptions built into the model are fundamentally wrong. This was the primary cause of the "2008 Financial Crisis," where models assumed that housing prices would never fall on a national scale. When that assumption failed, every "Prediction" built on top of it collapsed, leading to a global systemic failure. Another critical consideration is "Data Lag." Traditional financial statements are only released every three months. In a fast-moving crisis, a company can go from "Safe" to "Insolvent" in a matter of weeks. This is why "Market-Based Models" (like Merton) have become more popular, as they react to the daily fluctuations of the stock market. However, market-based models have their own flaw: "Market Noise." A stock price might crash because of a temporary panic or a "Short Squeeze," which doesn't necessarily mean the company is going to default. Finally, there is the human element of "Strategic Default." Sometimes, a borrower is capable of paying but *chooses* not to because the value of the collateral has dropped below the value of the loan. No mathematical model of "Solvency" can perfectly predict the "Behavioral" choice to walk away from a debt. Investors must always supplement quantitative predictions with "Qualitative Judgment," looking at the management's integrity and the broader geopolitical environment.