Stop Explaining Default, Just Price It: Jarrow and Turnbull's Reduced-Form Model

The structural credit models of Merton, Black and Cox are beautiful. They tell a story: a company defaults because the value of what it owns falls below the value of what it owes. Cause, then effect.

They also have an inconvenient property. To use one, you need to know the market value and the volatility of a company's assets, and nobody trades "all the assets of General Motors." You have to reverse-engineer those numbers from the stock price, which is fine in the classroom and irritating on a trading desk, where you have to quote a price on a credit derivative in the next thirty seconds and get the hedge right.

Robert Jarrow and Stuart Turnbull took a different route. Their 1995 paper essentially says: we do not need to know why the company defaults. We need to know how likely it is, and what we get back if it happens. Let the market tell us both.

The problem: the structural approach was too curious

Say a client wants to buy an option on a corporate bond, or a swap with a counterparty who might not survive to the final payment. To price it, you need default risk baked in.

The structural route says: build a model of the issuer's assets, estimate their volatility, find the barrier, compute the probability of hitting it. That is a research project, and it produces answers that are highly sensitive to inputs you cannot observe. Worse, the resulting model usually fails to reproduce the prices you can already see. If your model says a bond should yield 80 basis points over Treasuries and the bond is actually trading at 150, you cannot hedge anything with it, because your model and the market disagree about the very asset you are trying to trade around.

For a derivatives desk, that is backwards. The whole logic of derivatives pricing since Black and Scholes has been: take the prices of the traded things as given, and price the untraded thing so that no arbitrage is possible between them. Nobody models the "true" volatility of a stock from first principles before pricing an exotic option. They read the implied volatility off the market.

Jarrow and Turnbull applied that same discipline to credit.

The key idea via analogy: default as a lightning strike

Forget the story about assets falling below debts. Instead, treat default as a bolt of lightning.

Lightning does not have a cause you model in a pricing formula. It has a rate. In a thunderstorm, the chance of a strike in the next minute is higher. In clear weather, it is lower. You do not need the physics of charge separation to buy insurance. You need the strike rate.

That rate is what credit people call the default intensity or hazard rate: the probability that the company defaults in the next instant, given it has not defaulted yet. A rock-solid company has a low intensity. A wobbling one has a high intensity. Stack those instantaneous probabilities across time and you get a full curve of survival probabilities: how likely is this issuer to be alive in one year, two years, five.

That was the first ingredient. The second was recovery: if the lightning strikes, you do not lose everything. Bondholders typically recover some fraction of face value out of the wreckage. So the payoff on a defaultable bond is: the promised amount if the company survives, some recovery amount if it does not.

And here is the move that makes the whole thing click. Jarrow and Turnbull framed the risky payment using an analogy to foreign currency. Think of a dollar promised by a risky company as a payment in a foreign currency. It is written in dollars, but its value in real, certain dollars depends on an exchange rate, and that exchange rate is the market's assessment of whether you will actually get paid. When the company is healthy, the exchange rate is near one. When default hits, the rate collapses to the recovery value. Pricing a claim on a risky company becomes pricing a claim in a foreign currency, and the arbitrage-free machinery for currencies was already fully worked out.

Now flip the whole thing around, which is the practical payoff. You do not estimate the default intensity, you extract it. Corporate bonds trade. Their yields sit above government yields. That spread is the market's price for default risk. Feed the observed spreads into the model and solve backwards for the intensity curve that reproduces them. You now have a model that agrees with every price you can see, and you can use it to price the thing you cannot see: the option, the swap, the credit derivative.

The paper also dealt with a second, nastier form of credit risk that most people ignored at the time: the guy on the other side of your trade might also default. If you buy protection or hold a winning swap, its value to you depends on your counterparty actually paying. Jarrow and Turnbull priced that too, decades before counterparty risk became a household term in 2008.

Why it mattered

• It is the model the credit market actually runs on. The entire credit default swap market, the standard curve-building conventions, the way a trading desk bootstraps survival probabilities from quoted spreads: all of it is reduced-form pricing in this lineage. When a credit trader says "the five-year is at 200," the machine turning that into a hazard rate is Jarrow and Turnbull's descendant.
• It separated two questions that were being confused. "Why do companies default" is an economics question. "What is this credit-sensitive contract worth given what the market thinks" is a pricing question. Structural models answer the first. Reduced-form models answer the second. Trying to do both at once was making everyone worse at each.
• It made credit derivatives possible as a business. You cannot run a market-making book on a model that disagrees with observed prices. Calibration to market was the precondition for the credit derivatives market that exploded over the following decade.
• It got counterparty risk on the agenda early. The idea that the writer of a derivative, not just the underlying, carries default risk was a genuinely prescient inclusion.

The honest limitations

• It is a thermometer, not a diagnosis. The model tells you what the market thinks the default probability is. It has no view on whether the market is right. Calibration is not insight, and a model that always agrees with prices can never tell you a price is wrong.
• Default and recovery are tangled. The observed spread is roughly the default probability multiplied by the loss you take. Two unknowns, one equation. In practice people assume a recovery rate (forty percent is the industry's favourite napkin number) and back out the intensity, which means every default probability quoted in the market is partly an artefact of an assumption.
• Lightning strikes are not really random. Treating default as an exogenous bolt from the blue is convenient and false. Defaults cluster: they arrive in waves, in recessions, in whole industries at once. A basic intensity model with independent lightning badly underestimates the chance that many things fail together, which is precisely the risk that destroys portfolios, and precisely the risk that structured credit products of the 2000s were built to ignore.
• No economic story means no early warning. A structural model can tell you a company is running out of room. An intensity model can only tell you that spreads have widened, which you already knew from looking at the screen.

The one-line takeaway

Jarrow and Turnbull stopped asking why companies default and instead treated default as a random strike with a market-implied rate, calibrating credit risk directly from observable bond prices, which is why every credit derivatives desk in the world uses a version of their model rather than a structural one.