FRM® Merton distance to default: what is the threshold?

On this page

Merton distance to default measures how far a firm's modelled asset value lies above a debt payment threshold at a chosen future date, after allowing for asset volatility. A larger distance implies a smaller model probability of assets finishing below debt, if the model and its inputs are accepted. For FRM Part II, the decisive distinction is whether the calculation uses an actual expected asset return or a risk-free return: these produce different probability measures.

Begin with the balance sheet payoff

Imagine a company whose only promised payment is a single debt amount due at the end of the horizon. If its assets then exceed that promise, debtholders receive what they are owed and shareholders keep the residual. If assets fall short, shareholders' residual is zero and debtholders bear the shortfall. This payoff has the shape of a call option on the firm's assets with the debt promise as its exercise threshold. The option language is useful because it makes the default event explicit: the terminal asset value falls below the due amount. GARP's Merton discussion uses this structural interpretation. It is a model of a stylised capital structure, not an assertion that every real firm has one simple debt payment.

The difference between current assets and the debt face amount is only a starting cushion. A firm can appear comfortably above its obligation now yet have a substantial chance of crossing the threshold if its assets are volatile or the horizon is long. Conversely, a smaller present cushion may be less alarming under a calmer asset process. Distance to default puts the cushion, the expected path and the uncertainty into a common scale. That scale is often described as standard deviations, but the underlying expression uses log asset values. A candidate should therefore avoid subtracting two balance-sheet figures and calling the resulting currency amount a distance to default.

Read the variables before calculating

In the simple version, current asset value is V, the promised debt amount is D, the horizon in years is T and annual asset volatility is sigma. The expected annual asset return under the actual probability measure is mu. In a lognormal asset model, the physical distance to default is the log of V divided by D, plus the horizon-adjusted drift mu less half the variance, all divided by volatility times the square root of the horizon. The corresponding probability is the standard normal cumulative probability below the negative distance. The GARP-hosted default-model paper sets out both this expression and the separate risk-neutral calculation. The half-variance term belongs in the log process; omitting it changes the answer.

The distinction between physical and risk-neutral probability matters before any normal table is used. For the physical estimate, mu represents an assumed actual expected asset return. For the pricing measure, replace that drift with the risk-free rate; the resulting d2 gives a risk-neutral default probability. A price-implied probability is useful in valuing contingent claims, but it should not be presented as a forecast of observed default frequency without further modelling and calibration. The difference is about the probability measure, not a dispute over which normal table to use. If a question gives a drift, identify what it represents. If it supplies only a risk-free rate and asks for option-based valuation, use the pricing measure.

DD_P = \frac{\ln(V/D) + (\mu - \sigma_V^2/2)T}{\sigma_V\sqrt{T}}

PD_P = N(-DD_P)

Work an illustrative firm from threshold to probability

Take an invented firm with assets worth 100 monetary units today, one debt payment of 80 due in one year, annual asset volatility of 20%, an assumed actual asset drift of 5%, and a risk-free rate of zero. These are teaching inputs, not observed market data or GARP statistics. The current asset-to-debt ratio is 1.25. Its natural log is about 0.22314. The variance correction is half of 0.20 squared, or 0.02; the denominator is 0.20 because the square root of one year is one. With the actual drift, the numerator is 0.22314 plus 0.05 less 0.02, giving 0.25314. Dividing by 0.20 gives a physical distance of about 1.266.

The normal left-tail area below negative 1.266 is about 10.3%, so the simple model's physical default probability for this invented one-year horizon is about 10.3%. That figure is conditional on the assumed asset process, debt threshold and drift; it is not a measured default rate. If the same inputs are evaluated with the zero risk-free rate instead, the numerator becomes 0.20314 and d2 is about 1.016. The corresponding risk-neutral left-tail area is about 15.5%. The two answers are not arithmetic rivals. They answer different questions under different probability measures. A candidate who reports 15.5% as the firm's expected realised default frequency has silently changed the question.

It is helpful to test direction before relying on a calculator. Holding other inputs fixed in this example, a higher debt threshold narrows the asset cushion and lowers distance. Higher volatility spreads the terminal asset distribution and can raise the probability of crossing the threshold. A longer horizon changes both drift and uncertainty, so it requires recomputing the whole expression rather than multiplying a one-year probability. A higher assumed physical drift raises the physical distance in this particular setup, while it does not change a risk-neutral d2 calculated with an unchanged risk-free rate. These checks catch sign and measure errors that can hide behind plausible decimal answers.

Why the inputs are harder than the formula

A traded company's equity value may be visible, but total firm asset value and asset volatility are not quoted in the same direct way. The structural model treats equity as an option and uses the equity price and its volatility to infer those latent asset inputs. That inference is model-dependent. Equity is a levered residual claim, so substituting equity volatility directly for asset volatility tends to distort the distance. Likewise, adding an accounting debt balance to market equity may be a rough shortcut rather than a solved asset value. The FRM study-materials page locates credit risk measurement in Part II; for exam preparation, the important skill is to state what is observed, what is inferred and what the chosen model assumes.

Real debt has staggered maturities, covenants, collateral, revolving facilities and possible payments before the horizon. A single payment date cannot encode all of them. Default may also occur through liquidity pressure or contractual breach before assets are below one terminal face amount. Market prices may move abruptly, and volatility need not stay constant. These are substantive limitations because changing the effective threshold, horizon or process changes the computed distance. A precise-looking probability built from uncertain inputs is still conditional. When comparing firms, use consistent definitions of debt and time horizon and avoid treating a higher distance as a comprehensive credit rating.

Why the threshold definition matters

The threshold must match the debt obligation in the modelled horizon. If an exercise changes the debt amount from 80 to 90 while keeping the other illustrative inputs fixed, the log asset-to-debt cushion shrinks from the log of 1.25 to the log of roughly 1.11. The distance and the associated tail probability must therefore be recalculated; the earlier probability cannot simply be adjusted by ten percentage points. The same principle applies to a different horizon. Both the promised payment and the uncertainty over asset value may change. This is why the first line of a sound solution writes down which obligation is being tested and when it becomes due.

A useful conceptual test is to ask what would happen if volatility approached zero while today's assets remained safely above a fixed debt amount. The uncertainty around the terminal asset outcome would contract. Yet this intuition depends on the expected path and on the firm staying above the threshold over the stated horizon. It does not mean low historical equity volatility proves safety: the model needs asset volatility, and market conditions can change. Use such directional checks to audit the calculation, then return to the stated assumptions before drawing a conclusion about credit risk.

What to do with the result in an exam answer

First state the event: modelled assets are below the promised debt amount at the chosen horizon. Second name the probability measure. Third show the distance calculation, including the log ratio, drift, variance adjustment and volatility scaling. Fourth translate distance through the correct normal tail and label the probability as model-implied. Finally state one real limitation that relates to the case, such as debt structure or an inferred volatility. That sequence demonstrates understanding even when a question provides the asset inputs and does not ask you to solve the equity option equations.

A nearby but different credit-risk question concerns expected loss, which also needs loss severity and exposure. Distance to default alone supplies neither. If you need that separation, the FRM program overview places the concepts in their study context, while Omni's methodology page explains its approach to working through topics. Use the model here as a disciplined way to connect assets, debt and uncertainty; do not turn its output into an unqualified prediction or an investment decision.