The Mechanics Behind a Credit Spread
A credit surface starts with a contract, not with a graph. The focus here is the mechanism that turns uncertain collateral value into a loan amount, an expected loss, and a spread; understanding that chain lets readers see why the price of credit changes when a borrower promises more. The useful distinction is between what a borrower promises to repay and what a lender expects to recover or is willing to advance.
A promise meets an uncertain collateral value
The model in Credit Surfaces and Economic Uncertainty considers a contract made before the collateral’s future value is known. Let X denote the collateral value at the future date, and let j denote the amount the borrower promises to pay then. The promise may be attached to a secured bond or to a senior corporate obligation whose repayment is supported by the firm’s assets. The analysis holds other contract terms fixed so that the effect of changing the promise can be examined on its own.
The central enforcement rule is simple: the lender receives the smaller of the promise and the collateral value. If the collateral is worth at least j, the lender receives the full promise. If it is worth less, the lender can take the collateral but cannot collect the shortfall. The contractual payoff is therefore
min(j, X).
This is a no-recourse arrangement in the model. The lender has the right to seize the collateral, but no claim beyond it; the borrower loses the collateral in a default state but faces no additional penalty within the contract. The rule makes the source of credit risk visible. Risk does not come from a mysterious surcharge added after the fact. It comes from the possibility that future collateral value will fall below the promised repayment.
The timing is important. At the moment the contract is priced, X is not a number known to the lender; it is a random future outcome described by a distribution F. The lender values the contract by considering all the possible collateral states and the payoff delivered in each one. A promise is therefore a face amount, while the loan value is an expectation over possible recoveries. Treating those two quantities as identical would erase the very uncertainty that creates the spread.
The value of X is uncertain because it can take different values in different future states. A lender therefore does not price the bond from one chosen collateral value. The lender forms an expected payoff using a distribution F of possible values. The forward value of the loan is
L(j) = E[min(j, X)].
The notation matters. L(j) is the amount of loan value associated with a promise of j. It is not the promise itself. A borrower may promise 100 while the lender advances less than 100 in forward value because some future states will deliver less than the promise.
A small hypothetical example makes the difference concrete. Suppose collateral can be worth either 80 or 120 at the future date, with each outcome equally likely. For a promise of 100, the state-by-state payoffs are:
| Future collateral value | Contractual payoff |
|---|---|
| 80 | 80 |
| 120 | 100 |
The expected loan value is therefore 90. A promise of 70 is fully covered in both states, so the loan’s expected payoff is 70. A promise of 120 pays 80 in the low state and 120 in the high state, so its expected payoff is 100. The loan value has increased as the promise increased, but it has not increased dollar for dollar: the last 20 dollars of promised repayment add only 10 dollars of expected loan value.
That pattern is the basic reason the loan pricing function is increasing but concave. A larger promise cannot reduce the lender’s payoff, because the lender receives at least as much in every state. But the extra promised dollar is valuable only in states where the original promise is paid. As the promise moves higher, more states fall short, so fewer states contribute the full marginal dollar.
Concavity is a statement about the contract’s marginal exposure, not a claim that the borrower’s balance sheet is shrinking. The face amount j keeps rising. What changes is the conversion rate between that face amount and the expected loan value. Near a promise that is covered in almost every state, an increase in j looks nearly like an increase in the lender’s expected payoff. Near a promise that exceeds collateral in many states, the same increase mostly enlarges a shortfall rather than a recoverable payment.
The Federal Reserve paper expresses this intuition through the right derivative of the loan pricing function:
L′+(j) = 1 − F(j).
The quantity F(j) is the probability that collateral value is no greater than the promise, or the promise-insolvency probability in the paper’s terminology. Its complement is the probability that the collateral covers the promise. When the promise increases slightly, the loan value rises at the rate given by that solvency probability. If almost every state covers the old promise, a small increase in the promise adds nearly its full amount to expected loan value. If many states already fall short, the same increase adds much less.
This derivative is more than a technical convenience. It connects the local shape of a credit relationship to an economic event. The marginal value of a larger promise is the chance that the lender will actually be able to collect that additional amount. As leverage rises against the same collateral, the relevant solvency probability changes. The spread–leverage curve is therefore built from the distribution of possible collateral outcomes, state by state.
The diagram’s sequence also clarifies what is and is not being averaged. The lender averages the payoff after the collateral state is realized, not the collateral value first and the contract payoff afterward. Because the payoff is capped at j, a very high collateral outcome does not compensate the lender for a low outcome in the same way that an uncapped average might suggest. The contract’s state-by-state rule comes first; the expected loan value comes from averaging the resulting payoffs.
From expected payoff to a spread
The lender’s shortfall in any state is the difference between the promise and the amount actually received. In the model, the expected loss on a promise of j is
Λ(j) = E[j − min(j, X)].
The paper also writes this expected loss as the accumulated probability of collateral values below the promise:
Λ(j) = ∫₀ʲ F(x) dx.
That expression gives a useful visual interpretation. As the promise rises, the lender adds up the probability mass associated with collateral values below each level. The loan pricing function accumulates solvency probabilities; the expected-loss function accumulates insolvency probabilities. Together they split the promise:
L(j) + Λ(j) = j.
This identity separates three quantities that are often blurred in casual discussions of credit. The promise is the amount written into the contract. The loan value is the expected collateral-limited payoff. The expected loss is the gap between them. The spread is built from the last two quantities, not from the face amount alone. A promise can be large without producing a large spread if the collateral covers it across the relevant states; it becomes expensive when the expected shortfall grows relative to the value advanced.
Return to the hypothetical collateral values of 80 and 120. For a promise of 100, the lender receives 80 in one state and 100 in the other, so the expected loan value is 90 and the expected loss is 10. For a promise of 120, the expected loan value is 100 and the expected loss is 20. The promise has increased by 20, but the expected loss has also increased because the larger promise exposes the lender to a shortfall in the low-collateral state and a larger shortfall there.
The promise credit spread is the promise relative to the loan value, less one:
S(j) = j/L(j) − 1.
Using the identity above, the same expression can be written as
S(j) = Λ(j)/L(j).
In words, the spread is expected loss per unit of loan value. In the toy example, a promise of 100 produces a loan value of 90 and an expected loss of 10, so the spread is 10/90, or about 11.1 percent in the model’s forward-value units. A promise of 120 produces a loan value of 100 and an expected loss of 20, so its spread is 20 percent. These are not observed market quotations; they simply show how the formula works when the distribution and payoffs are explicit.
The denominator matters. Dividing by the promise would answer a different question: what fraction of the face amount is expected to be lost? Dividing by L(j) asks how much expected loss is attached to each unit of loan value supplied. The paper uses the latter ratio for the credit surface because the lender’s advance is the economic quantity being priced. As the collateral becomes less able to support a larger promise, both the numerator and denominator can change, and their ratio captures the resulting credit premium.
The example also shows why the spread rises as the promise becomes more demanding. At a promise of 70, the collateral covers the borrower in both possible states, so the expected loss is zero. At 100, one state falls short. At 120, the same low state falls even further short. The lender is not charging more because the word “120” is intrinsically expensive. The ratio of expected shortfall to expected loan value has changed.
That ratio is the bridge between the contract and the credit surface. If we vary j while keeping the collateral distribution and other contract terms fixed, we generate a sequence of loan values and spreads. The sequence can be plotted as a relationship rather than summarized by one number. Its shape depends on how the probability of shortfall accumulates as the promise moves through the possible values of X.
There are two reasons to keep promise size and loan value separate. First, they belong to different sides of the transaction. The borrower commits to j, while the lender values the repayment stream at L(j). Second, the same loan amount can require different promises under different collateral distributions. If uncertainty makes low collateral outcomes more likely, a borrower may need to offer a larger promise to raise the same expected loan value. The price of the loan and the amount promised are then related but not interchangeable.
This distinction is also why the paper later uses more than one leverage coordinate. A loan-to-value measure asks how large the loan is relative to the collateral. A promise-to-value measure asks how large the promised repayment is relative to the collateral. The present mechanism supplies both quantities, but it does not say that they will display the same curve. Choosing the coordinate changes which part of the contract is treated as the input and which part is treated as the response.
The model also states assumptions that keep the interpretation disciplined. The forward collateral and bond values are evaluated using a common pricing framework and discounting convention in the basic presentation. In a complete-market interpretation, the distribution can be read as the risk-neutral probability used to price the cash flows. In an incomplete-market interpretation, it can instead represent lenders’ probability assessments. The mathematics therefore identifies the uncertainty relevant to pricing under the model, not necessarily an objective forecast that every participant would agree on.
The mechanism can now be stated compactly. Future collateral value X determines whether a promise j is paid in full. The payoff rule produces the loan pricing function L(j). The gap between j and L(j) is expected loss. Dividing that loss by the loan value produces the spread. As j rises, the marginal loan value is governed by the probability of solvency, which is why the spread–leverage relationship carries information about the underlying distribution of collateral outcomes.
A spread is thus the endpoint of a chain, not the whole chain. To understand what the resulting curve can reveal, we need to reverse the mapping: start with the loan value, recover the promise that supports it, and ask what the complete relationship says about the uncertainty behind collateral.