FRM Exam Part II · Estimating Default Probabilities

Real-World vs Risk-Neutral Default Probabilities Explained

Updated 11 October 2026 · Fact-checked

Real-world (physical) default probabilities are estimated from historical default data. Risk-neutral probabilities are backed out of bond yields or CDS spreads. Risk-neutral values are usually higher because they include a risk premium and liquidity effects. Use risk-neutral PDs to value credit instruments; use real-world PDs for scenario analysis and loss forecasting.

Understand Real-World vs Risk-Neutral Default Probabilities

A real-world (physical) default probability is the actual chance a firm defaults. You estimate it from history, such as rating agency default rates, or from internal models. It answers: how likely is default, and how much will I really lose?

A risk-neutral default probability is the number that makes a pricing model reproduce today's market price. You get it from bond yield spreads or CDS spreads. A common approximation is hazard rate ≈ spread ÷ (1 − recovery rate). It answers: what default probability, in a world where investors ignore risk, gives this price?

Risk-neutral probabilities come out higher for several reasons. Investors are risk-averse and want compensation for bearing default risk. Defaults cluster in bad times, when losses hurt most, so investors demand extra return for that systematic risk. Bond spreads also include liquidity premiums, tax effects and, for CDS, supply and demand pressures. All of these push the implied PD above the historical one. The gap tends to be larger for high-rated names and in stressed markets.

The two are not rivals. They answer different questions. For valuing or pricing a credit instrument consistent with market prices (CDS, bonds, CVA), use risk-neutral PDs. For estimating actual losses, such as scenario analysis, stress testing, expected loss and capital, use real-world PDs. Mixing them up is the classic exam error.

Key formulas to remember

Average hazard rate from spread: λ ≈ s ÷ (1 − R); s = credit spread (annual, decimal), R = recovery rate. A risk-neutral estimate, approximate.
Cumulative default probability: Q(t) = 1 − e^(−λt); Assumes a constant hazard rate λ over t years.
Risk-neutral vs real-world relationship: λ(risk-neutral) = λ(real-world) × hazard rate ratio, usually ratio > 1; The ratio is the risk-premium effect; exceeds 1 in most markets. Not a fixed constant.
Spread decomposition: Spread ≈ expected loss component + risk premium + liquidity and other components; Only the expected loss part relates to real-world PD × (1 − R).
Expected loss rate (real-world): EL rate ≈ PD(real-world) × (1 − R); Spread minus this is the excess spread, a measure of compensation for risk.

How to solve Real-World vs Risk-Neutral Default Probabilities questions

Use this method for any question on real-world versus risk-neutral default probabilities.

1Identify the purpose: valuation or pricing consistent with market prices (risk-neutral) or loss estimation, scenarios and capital (real-world).
2Identify the data source: spreads, yields or CDS imply risk-neutral PDs; historical default rates or rating data give real-world PDs.
3If computing, convert spread to hazard rate with λ = s ÷ (1 − R). Keep spread as a decimal and use the stated recovery rate.
4Convert to cumulative probability with 1 − e^(−λt) if the question asks for a multi-year probability.
5For the real-world side, take the given historical PD, or the expected loss and divide by (1 − R).
6Compare: the risk-neutral value is normally larger. Explain the gap with risk aversion, systematic default risk and liquidity.
7State the conclusion for use: price with risk-neutral, forecast losses with real-world.

Quickest way: Purpose test and one-line hazard rate

When to use it: For conceptual MCQs and quick spread-to-PD conversions when time is short.

Ask: am I pricing or forecasting losses? Pricing means risk-neutral, forecasting means real-world.
If asked why they differ, pick the option mentioning risk premium, systematic risk or liquidity.
For numbers, compute s ÷ (1 − R) first; it is the annual hazard rate.
Check that your answer is a rate, not a cumulative probability, before choosing.

Common mistakes in Real-World vs Risk-Neutral Default Probabilities

Using risk-neutral PDs to forecast actual defaults or expected losses
Market-implied numbers look current and objective.
Fix: They overstate actual default likelihood. Use real-world PDs for loss forecasts, stress tests and capital.
Using real-world PDs to value a CDS or bond relative to market prices
Historical data feels more accurate.
Fix: Valuation consistent with traded prices needs risk-neutral inputs. Real-world PDs would produce prices that offer arbitrage against the market.
Saying the difference is only due to liquidity
Liquidity is the most familiar spread component.
Fix: Include the default risk premium for systematic risk and risk aversion as the main reason, with liquidity and tax as additional factors.
Forgetting to divide by (1 − R) when converting spread to hazard rate
Students treat the spread as the default probability.
Fix: Spread compensates for loss given default, so λ ≈ s ÷ (1 − R).
Assuming risk-neutral PD is always higher
It is true in most data, so it gets memorized as a law.
Fix: Say usually or typically higher. It is an empirical regularity, not a mathematical rule.

Worked examples

Example 1

A 5-year CDS on a corporate trades at a spread of 240 basis points. The recovery rate is 40%. The historical average default intensity for this rating is 1.2% per year. (a) Estimate the risk-neutral hazard rate. (b) Which is higher, and by what ratio?

Show the solution

Convert spread to decimal: 240 bp = 0.0240.
Risk-neutral hazard rate λ = 0.0240 ÷ (1 − 0.40) = 0.0240 ÷ 0.60 = 0.04, or 4.0% per year.
Real-world intensity is 1.2% per year.
Ratio = 4.0% ÷ 1.2% = 3.33.

Answer: The risk-neutral hazard rate is about 4.0% per year, about 3.3 times the 1.2% real-world rate. The gap reflects risk premium and other spread components.

Example 2

A bank holds a loan to a firm whose bond spread implies a risk-neutral PD of 4% per year (recovery 40%). The historical PD for the rating is 1.2%. The risk manager must (i) estimate expected annual credit loss for the loan book and (ii) mark a CDS on the same name. Which PD goes with each task? Find the expected loss rate for (i) on a ₹100 crore exposure.

Show the solution

Task (i) is loss estimation, so use the real-world PD of 1.2%.
Task (ii) is valuation consistent with market prices, so use the risk-neutral PD of 4%.
Expected loss rate = 1.2% × (1 − 0.40) = 0.012 × 0.60 = 0.0072, or 0.72%.
Expected loss on ₹100 crore = 0.0072 × ₹100 crore = ₹0.72 crore, or ₹72 lakh.

Answer: Use the real-world PD for expected loss: ₹72 lakh a year. Use the risk-neutral PD for marking the CDS. Using 4% for expected loss would give ₹2.4 crore, which overstates the loss.

Exam tips

Look for the verbs. Value, price, mark or hedge means risk-neutral. Forecast, scenario, stress or capital means real-world.
If an option says the ratio of risk-neutral to real-world PD is below 1 always, or equal to 1, reject it. Higher is the typical result.
Check units: basis points to decimals, and annual hazard rate versus cumulative probability.
In case questions, state both the measure and the reason, such as risk premium or liquidity, to match how options are worded.

Practice questions from Estimating Default Probabilities

Real-World vs Risk-Neutral Default Probabilities in other exams

The same ground in other exams, if you are preparing for more than one or want another angle on it.

Real-World vs Risk-Neutral Default Probabilities: frequently asked questions

Why are risk-neutral default probabilities higher than historical ones?

Market spreads pay for more than expected loss. They also compensate for risk aversion, the fact that defaults cluster in bad times, and liquidity and tax effects. Backing a PD out of the spread loads all of this into the implied number.

When should I use risk-neutral rather than real-world default probabilities?

Use risk-neutral PDs to value credit instruments, such as CDS, bonds and CVA, so prices match the market. Use real-world PDs for scenario analysis, stress testing, expected loss and capital estimation.

Is the risk-neutral PD a forecast of defaults?

No. It is a pricing input that reproduces market prices under a risk-neutral model. It tends to overstate actual default likelihood, so it should not be treated as a prediction.

How do I get a PD from a CDS spread in the exam?

Use the approximation hazard rate = spread ÷ (1 − recovery rate), then cumulative PD = 1 − e^(−λt). Convert basis points to decimals first.