Probability of Default: How Lenders and Bond Investors Assess Credit Risk

The Merton Model: Structural Approach to PD

The Merton model, developed by Nobel laureate Robert C. Merton in 1974, treats a company's equity as a call option on its assets with a strike price equal to the face value of its debt. The company defaults if the value of its assets falls below the value of its debt at the time of maturity. The model uses observable inputs — the company's market capitalization, debt levels, and equity volatility — to estimate the distance to default, which is then mapped to a PD using the normal distribution. A company with assets worth $500M, debt of $300M, and asset volatility of 25% might have a distance to default of 2.5 standard deviations, corresponding to a PD of approximately 0.6%. The Merton model is the foundation for commercial products like Moody's KMV, which refines the approach using proprietary databases of historical defaults. The model works best for publicly traded companies with liquid equity and simple capital structures. It struggles with private companies, financial institutions, and firms with complex debt structures. Corporate bond credit analysis →

Altman Z-Score: Accounting-Based Default Prediction

The Altman Z-score, developed by Edward Altman in 1968, uses financial ratios to predict bankruptcy probability within two years. The formula is: Z = 1.2A + 1.4B + 3.3C + 0.6D + 1.0E, where A is working capital / total assets, B is retained earnings / total assets, C is EBIT / total assets, D is market value of equity / book value of total liabilities, and E is sales / total assets. A Z-score above 3.0 indicates low bankruptcy risk; between 2.7 and 3.0 is a gray zone; between 1.8 and 2.7 is the distress zone; and below 1.8 indicates high bankruptcy probability. For a company with a Z-score of 1.5, Altman's original study found a 95% probability of bankruptcy within two years. The Z-score has been adapted for private firms (Z'-score) and service/manufacturing firms (Z''-score). It remains one of the most widely used bankruptcy prediction tools despite being developed over 50 years ago, though its predictive power diminishes for companies in industries with different financial characteristics than the original sample. High-yield bond default risk →

Credit Rating Based PD

Credit rating agencies publish historical default studies that map each rating category to average PDs over various time horizons. S&P's annual default study shows that AAA-rated issuers have a 0.01% one-year PD and 0.1% five-year PD, while B-rated issuers have a 3% one-year PD and 20% five-year PD. These ratings-based PDs are the most commonly used approach in the bond market because they are simple, transparent, and readily available. However, they have significant limitations: ratings are slow to change (they lag market pricing of credit risk), they are ordinal rather than cardinal (same rating does not mean same PD across industries or time periods), and they reflect through-the-cycle views rather than point-in-time risk. For these reasons, sophisticated investors use ratings as a starting point and adjust PD estimates based on current market conditions, sector trends, and company-specific analysis. The cumulative PD over time follows an exponential pattern — the longer the horizon, the higher the probability of default, but the relationship is not linear because companies that survive the first few years tend to be stronger credits.

Market-Based PD: Credit Spreads and CDS

Credit spreads — the yield difference between a corporate bond and a comparable-maturity Treasury — provide a market-implied PD that often leads rating agency assessments. The logic is that investors demand higher yields for higher default risk. A bond trading at a 300 basis point spread over Treasuries with an assumed recovery rate of 40% implies a PD of approximately [spread / (1 - recovery rate)] = 3% / 0.6 = 5% annualized. However, credit spreads also contain liquidity premiums, tax considerations, and risk premiums that vary over time. Credit default swaps (CDS) provide a purer measure of credit risk because they are standardized contracts referencing the same issuer. The CDS spread directly reflects the cost of insuring against default. A CDS trading at 200 basis points with a 40% recovery assumption implies a PD of approximately 200bps / (1 - 0.4) = 333bps, or 3.33% annualized. Market-implied PDs react quickly to news and changing conditions, making them more current than ratings-based PDs, but they also embed investor sentiment and risk appetite that can cause PD estimates to deviate from fundamental credit quality. Convertible bonds and credit risk →

PD in Expected Loss Calculations

Expected loss (EL) is the product of PD, LGD, and EAD. For a $5M loan with a 2% annual PD and 50% LGD, the annual expected loss is $5M x 2% x 50% = $50,000. Over a 5-year maturity, the cumulative expected loss is approximately $250,000, though the exact calculation requires accounting for the changing PD over time (the probability of surviving each year multiplied by the probability of defaulting in that year). Banks use this framework to set loan loss provisions under the Current Expected Credit Loss (CECL) standard, which requires estimating expected credit losses over the full life of the loan rather than recognizing losses only when they become probable. The PD term structure — how PD changes with the time horizon — is critical for long-duration assets. A 10-year bond does not have a PD equal to 10 times the one-year PD because the probability of surviving the first few years reduces the cumulative default probability. Understanding the relationship between PD, LGD, and EAD enables investors to calculate whether a bond's yield spread adequately compensates for credit risk.

What is the difference between point-in-time and through-the-cycle PD?

Point-in-time (PIT) PD reflects the current economic and company-specific conditions, making it higher during recessions and lower during expansions. Through-the-cycle (TTC) PD averages over the full economic cycle, smoothing out cyclical fluctuations. Credit rating agencies use TTC PDs — a BBB rating remains BBB through the cycle even as the company's actual default probability fluctuates. Banks use PIT PDs for loan pricing and provisioning because they reflect current risk. The difference between PIT and TTC PD can be substantial: a BBB-rated company might have a PIT PD of 0.5% during a recession and 0.1% during an expansion, while its TTC PD stays at 0.2%. Regulatory capital frameworks typically use TTC PDs for risk weighting, while accounting standards (IFRS 9, CECL) require PIT PDs for loan loss provisions. Investors should understand which type of PD they are looking at when assessing credit risk.

How do macroeconomic factors affect probability of default?

Macroeconomic conditions are the single largest driver of systematic default risk. GDP growth, unemployment rates, interest rates, and industry conditions all significantly affect PD. S&P's research shows that default rates during recessions are 3-5 times higher than during expansions across all rating categories. The most vulnerable companies are those with high leverage, low interest coverage, and weak competitive positions. Macroeconomic stress testing — simulating PDs under recession scenarios where GDP contracts, unemployment rises, and interest rates change — is a standard practice for credit portfolio management. The relationship between macro variables and PD is often modeled using logistic regression or machine learning techniques, with historical data showing that the unemployment rate alone explains a significant portion of aggregate default rate variation. Investors should always assess credit risk in the context of the current and expected macroeconomic environment, not just the company's standalone financial health.

What is the relationship between PD and recovery rates?

PD and recovery rates are negatively correlated — when defaults are high, recovery rates tend to be low. During systemic crises like 2008, many companies default simultaneously, flooding the market with distressed assets and depressing recovery values. Moody's data shows that the average recovery rate on senior unsecured bonds during normal periods is approximately 40%, but during crisis periods it falls to 20-30%. This negative correlation amplifies credit losses: when PD is high, LGD is also high, producing larger-than-expected losses. This phenomenon is captured in the concept of "downturn LGD" used in regulatory capital calculations, which requires banks to use LGD estimates that reflect economic downturn conditions rather than long-term averages. The PD-LGD correlation is a key source of credit portfolio risk that simple expected loss calculations may underestimate. Comparing high-yield and investment-grade →

How do lenders use PD to set interest rates?

Lenders use PD as a direct input to loan pricing. The interest rate on a loan can be decomposed into: risk-free rate (the cost of capital, typically a Treasury yield), expected loss component (PD x LGD, covering anticipated defaults), unexpected loss component (capital charge for unexpected losses, typically 2-3 times expected loss), operating costs (origination, servicing, monitoring), and profit margin. For a personal loan with a 5% annual PD and 70% LGD, the expected loss component is 3.5%. Adding the risk-free rate of 5%, a capital charge of 2%, operating costs of 1%, and profit of 1.5% produces a total interest rate of 13%. This risk-based pricing ensures that borrowers with higher PD pay higher rates, compensating lenders for the additional risk. Credit card APRs often exceed 20% because the PD for unsecured revolving credit is significantly higher than for secured installment loans. How credit scores relate to PD →

Related Resources