The One Trick That Made Credit Pricing Easy: Duffie and Singleton's Adjusted Discount Rate

By the late 1990s the reduced-form approach to credit was established: treat default as a random event with an intensity, calibrate that intensity from market prices, price your derivative. The trouble was that actually doing it was a mess.

Interest rates move. Default risk moves. They are correlated with each other. Recovery in default is uncertain. Every one of those pieces needs modelling, and once you stack them the mathematics gets ugly fast, with the default event and the discounting tangled together in a way that resists clean solutions.

Darrell Duffie and Kenneth Singleton's 1999 paper is beloved by practitioners because it found the modelling assumption that makes the whole knot fall apart.

The problem: default and discounting would not separate

To value any future cash flow you do two things: figure out what you expect to receive, and discount it back to today at the appropriate rate.

For a government bond, that is easy. You will receive the coupon. Discount it at the risk-free rate. Done.

For a corporate bond, both halves get complicated at once. What you receive depends on whether the company survives, which depends on a default process. The rate you discount at depends on interest rates, which are also moving randomly, and which may be correlated with whether the company survives, since recessions hurt both. And if default happens, you get a recovery amount that is itself uncertain and might depend on when the default occurred and what the bond was worth at the time.

Every earlier attempt to handle this produced expressions where the survival probability, the recovery, and the discount factor were all entangled inside the same expectation, and you could not compute one without dragging along the others. It worked, but it was slow and inflexible, and it did not slot into the enormous, well-oiled machinery that quants had already built for pricing ordinary interest rate products.

The key idea via analogy: fold the risk into the interest rate

Here is the intuition. Imagine you are lending money to a friend who might vanish. You could carefully model the probability of his vanishing, the fraction of the money you would claw back, and discount everything separately.

Or you could just say: "Lending to him is like lending to a bank, except at a worse interest rate. Let me figure out what worse rate makes the numbers come out right, and then treat him as a safe borrower at that rate."

That is Duffie and Singleton's move, made rigorous.

The key is a specific and clever choice about recovery. Rather than assuming you recover a fraction of the bond's face value, which is how lawyers think about it, they assume you recover a fraction of what the bond was worth just before it defaulted. This convention became known as recovery of market value. It sounds like a minor bookkeeping distinction. It is not. It is the whole trick.

Why? Because if default costs you a fixed fraction of the bond's current value, then the loss is always proportional to the value itself. And a risk of losing a constant fraction of value per unit of time is mathematically indistinguishable from... a slightly higher interest rate eating away at that value.

So you can define an adjusted discount rate: the risk-free rate, plus the default intensity multiplied by the fraction you would lose. Discount the bond's promised cash flows at that rate, exactly as if the bond were riskless, and you get the right answer. Default has been completely absorbed into the discounting.

The consequence is enormous. Every technique, model and piece of code that quants had spent twenty years building for the term structure of riskless interest rates now works, unchanged, on risky bonds. All the tractable models with clean closed-form solutions, all the calibration routines, all the pricing engines. You swap in the adjusted rate and everything runs.

Duffie and Singleton went on to show how to use this to build models where the credit spread has its own dynamics, mean-reverting and correlated with the level of interest rates, and to price things like options on credit spreads. But the reason the paper is famous is the trick: risky bond, safe machinery, adjusted rate.

Why it mattered

• It made credit modelling industrial. Before this, pricing a portfolio of risky bonds and credit derivatives meant bespoke, slow, fragile numerics. After it, credit desks could reuse the entire well-tested interest-rate toolkit. That is the difference between a research paper and a production system.
• It defined how spread curves are built. The mental model that a credit spread is a rate, that it has a term structure, that it mean-reverts, that it can be modelled with the same families of processes used for interest rates, is the practical language of every credit trading desk. This paper is where it was formalised.
• It let credit and rates be modelled together. Because both live inside the same discount rate, you can let them be correlated, which matters enormously: credit spreads tend to blow out at exactly the moments central banks are cutting rates.
• It powered the structured credit era, for better and worse. The tractability this paper delivered is part of what made it feasible to price and risk-manage the vast pools of credit derivatives that grew through the 2000s. The tools were sound. What people did with them was another matter.

The honest limitations

• The recovery assumption is a convenience, not a fact. Bankruptcy courts do not pay you a fraction of the bond's market value the moment before default. They pay out of an estate, in a process taking years, according to a legal priority order, with lawyers taking a cut. "Recovery of market value" is chosen because it makes the mathematics collapse, and it is at best a rough stand-in for what actually happens in a restructuring.
• You still cannot separate default from recovery. The model gives you the product of the default intensity and the loss fraction, because that product is what the spread reveals. To get one you must assume the other. The market's convention of assuming forty percent recovery is baked into an astonishing number of published default probabilities.
• The spread is not all credit. The adjusted rate absorbs everything that makes a corporate bond yield more than a Treasury, and that includes illiquidity, taxes, and plain supply-and-demand imbalance. Calling the whole thing "default risk" overstates how much default the market is really pricing, a point that later papers hammered home.
• Correlated defaults are still hard. Extending the framework to many issuers who might fail together, which is the risk that actually matters for a portfolio, remains the genuinely difficult part, and no amount of clever discounting makes it go away.

The one-line takeaway

Duffie and Singleton showed that if you assume default costs you a fixed fraction of a bond's current value, then all the default risk can be folded into a single adjusted discount rate, letting quants price risky bonds with the exact same machinery they already used for safe ones. That one assumption is why modern credit desks can run at all.