How to Calculate Probability of Default: Z-Score, Merton, and Logistic Models

To calculate the probability of default, choose a method that fits the borrower you’re assessing, gather the specific financial or credit data that method requires, and run the inputs through the model to produce a one-year default likelihood. For public corporates, that usually means the Altman Z-Score or the Merton distance-to-default framework. For individual consumers, logistic regression on credit bureau and account data. For portfolios of rated borrowers, historical transition matrices. Each produces a number, but the number only means what the method’s assumptions let it mean, so the choice of model is itself part of the calculation.

Gather the Right Inputs First

The data you need depends entirely on who you’re rating.

For a corporate borrower, start with the 10-K annual report and 10-Q quarterly filings submitted to the SEC.1Investor.gov. Form 10-Q Pull total assets, current assets, current liabilities (current assets minus current liabilities gives you working capital), retained earnings, total debt at face value, and market value of equity from the balance sheet. From the income statement, take earnings before interest and taxes (EBIT) and net revenue. If you plan to use a market-based model like Merton, you also need the historical volatility of the firm’s asset value, which analysts typically infer from stock price movements.

For a consumer borrower, the raw material comes from credit bureau reports and the bank’s own records. Payment history, outstanding balances, credit utilization, length of credit history, and public records like bankruptcies and liens come from the bureaus. Banks add their own data on income, employment status, time at current address, and the existing customer relationship.

The Altman Z-Score for Corporates

The Z-Score is the most widely taught statistical method for predicting corporate default. It combines five ratios, each with a fixed weight, into one composite score:

Z = 1.2(Working Capital / Total Assets) + 1.4(Retained Earnings / Total Assets) + 3.3(EBIT / Total Assets) + 0.6(Market Value of Equity / Total Liabilities) + 1.0(Sales / Total Assets)

The heaviest weight, 3.3, sits on EBIT-to-assets because operating profitability relative to the asset base is the strongest short-term predictor of financial collapse. Retained earnings over total assets captures cumulative profitability across the firm’s life, which is why younger firms tend to score lower even when currently profitable. Working capital over total assets measures short-term liquidity. Market value of equity over total liabilities reflects how much the market values the firm’s equity against what it owes.

Reading the output is simple. Above 2.99 is the safe zone: default within a year is unlikely. Between 1.81 and 2.99 is the grey zone, where the signal is ambiguous and further analysis is warranted. Below 1.81 is the distress zone, which flags a high probability of default or bankruptcy.

The model was originally calibrated on publicly traded U.S. manufacturing firms. For private companies, financial institutions, or firms in emerging markets, analysts typically use modified versions with recalibrated weights and cutoffs. The strength of the Z-Score is that you can produce one with a balance sheet and a stock quote. Its weakness is that it’s a static snapshot: it won’t capture deterioration between reporting dates, and the fixed weights don’t adjust to macroeconomic shifts.

The Merton Model and Distance to Default

The Merton model treats a company’s equity as a call option on its assets. Shareholders have the right, not the obligation, to pay off debt at maturity. If assets exceed debt at that point, shareholders keep the difference. If assets fall short, shareholders walk away and the firm defaults.2University of Toronto / Rotman School of Management. Merton’s Model, Credit Risk, and Volatility Skews Default becomes a boundary problem rather than a ratio test.

The key output is the distance to default (DD), which measures how many standard deviations the firm’s asset value sits above the debt threshold:

DD = [ln(V/D) + (μ − 0.5σ²) × T] / (σ × √T)

V is current market value of assets, D is face value of debt, μ is expected return on assets, σ is asset volatility, and T is the horizon, usually one year. A larger DD means more cushion before assets reach the default boundary. To turn DD into a probability, run it through the cumulative normal distribution: PD = N(−DD).2University of Toronto / Rotman School of Management. Merton’s Model, Credit Risk, and Volatility Skews

The practical catch is that asset value and asset volatility aren’t directly observable. Both must be inferred from equity prices and equity volatility using the Black-Scholes option pricing relationship, which introduces estimation error. Highly leveraged firms with volatile stock prices show a short distance to default; stable firms with low debt show a wide buffer. Because the inputs are market data, the model updates continuously, which makes it far more responsive to real-time deterioration than ratio-based methods. Moody’s commercialized this framework as its EDF (Expected Default Frequency) metric, which forecasts one-year default likelihood and treats missed payments, bankruptcy, and distressed restructurings all as default events.3Moody’s. US Firms’ Default Risk Hits 9.2%, a Post-Financial Crisis High

Logistic Regression for Consumer Borrowers

For retail portfolios, including credit cards, auto loans, and mortgages, logistic regression is the standard model. It’s designed around the data available on individuals: demographic characteristics, payment behavior, and account metrics.

The model estimates default probability as:

PD = 1 / (1 + e^(−z))

Here z is a linear combination of borrower characteristics. Typical predictors include age, income, time at current address, employment status, credit utilization rate, number of existing credit accounts, and account balance. Fitting the model to historical data produces a coefficient for each variable, and the logistic function compresses the weighted sum into a value between 0 and 1.

Logistic regression stays popular because its coefficients are interpretable. Each one tells you the direction and size of a variable’s effect on default probability, which matters when a regulator asks why a particular factor is in the model. If a variable turns out to be statistically insignificant or raises fair lending concerns, you can drop it and refit. The output probabilities map directly to score bands, which is how most consumer credit scores are structured: the score corresponds to a predicted PD for that band.

Transition Matrices for Rated Pools

Rating migration analysis is backward-looking. Instead of modeling individual financials, it tracks how credit ratings for pools of borrowers have shifted over time. Rating agencies and internal bank systems maintain transition matrices that show, for any starting rating, the percentage of borrowers that migrated to each possible rating, including default, over a one-year period.

If historical data shows 3.5% of B-rated corporate borrowers defaulted within one year over the last twenty years, that 3.5% becomes the baseline PD for current B-rated exposures. The matrix also reveals intermediate migration patterns, and those probabilities can be chained to estimate multi-year cumulative default rates.

The assumption is that past rating behavior predicts future behavior within the same grade. That holds reasonably well in stable conditions but can break down during systemic crises, when correlations spike and migration accelerates past historical norms. Migration analysis works best alongside a forward-looking model rather than on its own.

Point-in-Time or Through-the-Cycle

One decision changes the number materially: whether your PD is point-in-time (PIT) or through-the-cycle (TTC).

A through-the-cycle estimate smooths out short-term economic fluctuations. It asks what PD would be under average conditions across a full business cycle, regardless of where the economy sits today. TTC PDs are stable over time, which makes them useful for long-term capital planning. This was the approach favored under earlier Basel II rules.

A point-in-time estimate uses current and forecast macroeconomic conditions. It brings in variables like GDP growth, unemployment, and market indices alongside borrower-specific factors. When conditions deteriorate, PIT PDs rise; when they improve, PIT PDs fall. The Federal Reserve’s annual stress test scenarios illustrate how severe adverse projections push PIT estimates sharply higher.4Federal Reserve Board Publication. 2026 Stress Test Scenarios

Current accounting standards, CECL in the United States and IFRS 9 internationally, now require point-in-time projections that incorporate forward-looking information. CECL specifically requires banks to estimate lifetime expected credit losses using reasonable and supportable forecasts, not just historical averages.5U.S. Department of the Treasury. The Current Expected Credit Loss Accounting Standard and Financial Institution Regulatory Capital Study One practical move: build a PIT model with macroeconomic variables, then substitute long-run averages for those variables to derive a TTC estimate from the same framework.

Basel Requirements if the Number Drives Capital

Banks using the Internal Ratings-Based (IRB) approach must produce their own PD estimates under specific rules.6Bank for International Settlements. IRB Approach: Minimum Requirements to Use IRB Approach These aren’t suggestions; they determine how much capital the bank must hold.

  • At least seven borrower grades for non-defaulted exposures, plus one for defaulted borrowers.
  • PD estimates must represent the long-run average of one-year default rates for borrowers in each grade, calculated as a simple count-weighted average across obligors.
  • The historical data used must span at least five years.
  • A margin of conservatism must be added to account for the likely range of estimation errors.
  • No PD for a non-defaulted borrower may fall below 0.05% (5 basis points) for most exposure classes.

The resulting PD feeds the risk-weighted asset formula, which multiplies the capital requirement by 12.5 times exposure at default.7Bank for International Settlements. CRE32 – IRB Approach: Risk Components Higher PDs produce higher risk-weighted assets and require more capital. The minimum Common Equity Tier 1 ratio is 4.5% of risk-weighted assets, so underestimating PD across a portfolio means both mispricing loans and holding too little capital.

Validate the Number You Produced

Running the formula gives you a figure. Confirming it holds up is a separate step. Banks must compare realized default rates against their estimated PDs for each rating grade and show that outcomes fall within the expected range, with back-testing updated at least annually.6Bank for International Settlements. IRB Approach: Minimum Requirements to Use IRB Approach When realized defaults consistently exceed estimates, estimates must be revised upward.

External benchmarks add a second check. Comparing internal PDs against default rates published by rating agencies for similarly rated borrowers helps surface systematic bias. If your model puts a BB-rated portfolio at 1.5% PD and historical agency data shows BB defaults averaging 2.5%, that gap needs an explanation. Better credit selection is one possibility. A miscalibrated model is another, and the burden is on you to tell them apart.

Market conditions move, borrower financials shift, and macro forecasts turn over quarterly. A PD calculated in January from December statements and a benign outlook can look very different by June after a softer labor market and a missed covenant test. Institutions that recalibrate often and stress their portfolios under adverse scenarios are the ones whose PD estimates still mean something when conditions change.