Jarrow-Turnbull Model

July 30, 2026 Risk Management

The Jarrow-Turnbull model is a reduced-form framework for pricing defaultable securities and credit derivatives using default timing and recovery assumptions.

On this page

The Jarrow-Turnbull model is a reduced-form credit-risk framework for pricing defaultable securities and derivatives. Instead of deriving default from a firm’s asset value and capital structure, it models default as an event with a specified risk-neutral arrival process and combines that process with interest rates and recovery assumptions.

Key Takeaways

Default timing is modeled directly rather than triggered by an observable firm-value boundary.
Default intensity describes the conditional arrival rate of default, not a guaranteed frequency.
Market-calibrated intensity is generally a pricing measure that can include risk premia, liquidity, and technical effects.
Recovery convention, interest-rate dynamics, calibration instruments, and default-event definitions materially affect value.
A reduced-form pricing output should not be relabeled as a real-world probability of default without a justified conversion and validation.

Reduced-Form Credit Modeling

Structural models ask whether firm asset value falls below a debt boundary. Reduced-form models instead specify how default time (\tau) arrives. This permits unexpected default and can fit market prices without estimating an unobservable firm asset value.

The original Jarrow-Turnbull framework developed a methodology for pricing and hedging derivatives exposed to default by an underlying security or by the derivative writer. Modern intensity-model implementations vary and should not all be described as identical to the original paper.

Default Intensity and Survival

Let (\lambda(t)) denote a deterministic default intensity for illustration. Conditional survival to time (T) is:

Q(\tau > T) = \exp\left(-\int_0^T \lambda(u)\,du\right)

The cumulative default probability is:

Q(\tau \le T) = 1 - \exp\left(-\int_0^T \lambda(u)\,du\right)

If intensity is constant:

Q(\tau \le T) = 1 - e^{-\lambda T}

These equations are conditional on the intensity model and probability measure. If intensity is stochastic, correlated with rates, or state-dependent, valuation requires the appropriate expectation rather than simply inserting an average intensity.

Worked Intensity Example

Assume a constant annual risk-neutral intensity of 2.5% for three years:

Q(\tau \le 3) = 1 - e^{-0.025 \times 3} \approx 7.23\%

The result is not (2.5% \times 3 = 7.5%) because survival compounds continuously in this simplified setup.

If the assumed recovery rate is 40%, loss given default is:

\text{LGD} = 1 - 0.40 = 60\%

Under a highly simplified constant-intensity approximation, the credit spread is:

s \approx \lambda(1-R) = 0.025 \times 0.60 = 1.50\%

That is approximately 150 basis points. The approximation omits discounting details, accrued payments, liquidity, risk premia beyond the fitted intensity, contract terms, counterparty risk, and the selected recovery convention.

Recovery Conventions

Pricing can differ depending on what is recovered and when. Common modeling choices include recovery of:

a fraction of face or par value;
a fraction of the default-free value;
a fraction of market value immediately before default;
a contractually specified amount at default or later settlement.

A 40% recovery assumption is incomplete unless the model states the recovery base, timing, seniority, and treatment of accrued amounts.

Calibration

Reduced-form models can be calibrated to prices or spreads of:

corporate or sovereign bonds;
Credit Default Swaps;
other defaultable claims with suitable liquidity and terms.

Calibration should align:

reference entity and obligation;
seniority and restructuring terms;
currency and maturity;
recovery assumptions;
discount curve;
coupon and accrual treatment;
liquidity and data quality;
observation time and quote conventions.

An exact fit to a small set of market prices does not prove that intensity, recovery, or extrapolated probabilities are correct.

Risk-Neutral vs. Real-World Default Probability

Pricing models usually operate under a risk-neutral measure so discounted model prices match market prices. A risk-neutral default probability reflects pricing and risk premia under model assumptions. A real-world or physical PD aims to estimate observed default frequency for forecasting, underwriting, or risk measurement.

The two can differ because investors require compensation for systematic credit risk, liquidity, uncertainty, and other factors. The Basel Framework explicitly cautions against using market-implied PDs as objective PD estimates without correction and evidence.

Jarrow-Turnbull vs. Structural Models

Feature	Jarrow-Turnbull / reduced form	Merton / structural
Default trigger	Modeled arrival process	Firm asset value below debt boundary
Main calibration	Market prices, spreads, rates, recovery	Equity value, volatility, liabilities, rates
Unexpected default	Can occur by construction	Basic Merton default occurs at horizon
Capital-structure link	Indirect	Explicit
Typical strength	Flexible market pricing	Economic equity-credit relationship
Typical weakness	Intensity and recovery identification	Unobservable assets and restrictive structure

See Structural Model of Credit Risk for the family comparison.

Practical Uses

Reduced-form models can support:

Bond Valuation for defaultable claims;
credit-derivative valuation and hedge sensitivities;
term structures of survival and default probability;
scenario analysis for intensity and recovery;
counterparty-risk and valuation-adjustment frameworks;
relative-value comparisons across instruments.

The model does not replace legal review of credit events, deliverable obligations, settlement, netting, collateral, and counterparty exposure.

Limitations and Model Risk

Intensity and recovery can be difficult to identify separately from market prices.
Sparse or illiquid quotes can produce unstable calibration.
Market spreads include liquidity and technical effects.
Recovery is uncertain and correlated with default conditions.
Default and interest rates may be correlated.
Piecewise calibration can fit observed maturities but behave poorly between or beyond them.
A single-entity model may omit common factors and default dependence.
Counterparty and underlying-credit default can interact.
Risk-neutral outputs can be misused as forecasting probabilities.

Common Mistakes

Defining the model as a linear regression of intensity on interest rates.
Multiplying annual intensity by years and calling the result exact.
Omitting the recovery convention.
Treating a fitted spread as pure expected credit loss.
Mixing quotes with different seniority, currency, or documentation.
Extrapolating beyond liquid maturities without sensitivity analysis.
Using the pricing measure for underwriting or reserves without conversion and validation.

Model Governance

Document model purpose, pricing measure, data sources, calibration hierarchy, recovery convention, curve construction, numerical implementation, sensitivities, benchmark models, independent validation, monitoring, overrides, and conditions requiring recalibration or withdrawal.

Primary and Official References

Structural Model of Credit Risk: A model family that derives default from the relationship between firm asset value and obligations.
Merton Model: The foundational option-based structural model used as a contrast to reduced-form intensity models.
Credit Spread: A market input that reflects credit risk together with liquidity, risk premia, and other pricing effects.
Credit Default Swap (CDS): A credit derivative whose quotes can help calibrate default-intensity and recovery assumptions.
Probability of Default (PD): A default likelihood that must be labeled by horizon, definition, and probability measure.

Educational Use

This article is educational and does not provide individualized investment, trading, valuation, capital, accounting, or regulatory advice. Reduced-form model results depend on market data, probability measure, recovery, calibration, documentation, and intended use.

Corporate Failure Prediction

Merton Model