Credit Spread Is Not a Default Probability

Define the comparison before interpreting the gap

A simple nominal spread subtracts a benchmark yield from a bond yield. Both yields need compatible units and conventions for that subtraction to be meaningful. Matching only maturity may leave different coupon patterns and sensitivities. A spread to a nearby benchmark bond, a spread to an interpolated government curve and a constant spread over spot rates are related descriptions, but they are not identical measurements.

For an arithmetic illustration, assign annual yields of 5.2% and 3.7% under the same compounding convention. The difference is 1.5 percentage points, or 150 basis points. These are hypothetical comparison inputs, not quotes. Nothing in that subtraction specifies whether a payment will be missed, when it might be missed, how much could be recovered or what compensation is required for bearing uncertainty.

Separate compensation from expected loss

BIS research on sovereign spreads distinguishes expected default loss from a risk premium. Its useful conceptual lesson is that the price of bearing risk need not equal the average loss. A holder may require compensation for losses occurring in especially difficult conditions. Market liquidity and features of the chosen benchmark can also affect a measured yield gap without a matching change in the issuer's payment prospects.

This means that even a carefully measured spread is not a clean statistical estimate of default frequency. Reverse-engineering a probability requires a pricing model and additional assumptions. A probability inferred under a pricing measure also need not equal a real-world frequency forecast. Describing both as a default probability without naming the measure can make two different analytical questions appear to have the same numerical answer.

Make the expected-loss arithmetic explicit

Consider a separate one-year teaching example with USD 100,000 of exposure, an assumed 2% probability of default over that year and recovery of 40% of exposure if default occurs. Ignore coupons, discounting, timing within the year and changes in exposure. Loss given default is therefore 60%, and expected loss is 100,000 * 0.02 * 0.60 = USD 1,200.

An independent outcome check gives zero loss with 98% probability and USD 60,000 loss with 2% probability, again averaging USD 1,200. That average is 1.2% of the stated exposure. It is not a forecast that this single position will lose precisely USD 1,200, and it is not automatically an annual spread of 120 basis points. The example does not price risk or account for liquidity.

Keep horizon and recovery definitions attached

A one-year probability and a cumulative probability to maturity describe different events. Extending the former over several years requires assumptions about survival and changing conditions. Similarly, recovery can be measured against principal, an outstanding claim or another stated base. An analysis that copies a recovery percentage while changing its denominator may be internally inconsistent even when every formula is typed correctly and the output looks plausible.

Recovery timing also matters. Receiving an amount promptly and receiving the same amount after a lengthy resolution are not equivalent in present-value terms. Seniority, security and claim-specific provisions can change the outcome. These observations do not require an estimate for a particular issuer; they explain why a single spread field cannot replace a documented default and recovery framework when the question concerns actual payment impairment.

Distinguish repricing from a payment failure

Investor.gov identifies credit risk as the possibility that an issuer fails to make interest or principal payments on time. Spread widening can occur before any such failure, and a bond can continue paying while its modeled or traded value falls. Conversely, an actual default can produce a discontinuous loss that a small-change spread-duration approximation does not describe well. The two risks should not share an unlabeled result column.

For a review worksheet, separate a contractual-payment schedule, a no-default repricing scenario and a default outcome scenario. The first states promises; the second changes discounting while preserving those promises; the third changes what is received. Adding an assumed default loss on top of a spread-adjusted expected-loss valuation without inspecting its construction can count the same impairment twice. The treatment must be explicit before combining scenarios.

Check what changed on the benchmark side

A narrowing spread need not mean that the bond yield fell. In the earlier arithmetic, keep the assigned bond yield at 5.2% but raise the comparison yield from 3.7% to 4.2%. The gap becomes 1.0 percentage point, or 100 basis points. It narrows by 50 basis points even though the bond's own yield is unchanged. This is a decomposition, not a market scenario forecast.

The comparison suggests a useful question for any spread commentary: how much of the movement came from each side? Also ask whether the benchmark definition, maturity matching or observation time changed. Without that information, language such as improving credit may go beyond the evidence. A relative-price movement can motivate further investigation while remaining insufficient to establish a change in underlying default likelihood or expected recovery.

State what the educational calculator leaves out

The parent present-value tools assume the entered fixed coupons and bullet principal are paid. Changing the yield can illustrate a discount-rate effect, but the tools do not estimate survival probabilities, recovery cash flows or an issuer's capacity to pay. A higher entered yield is therefore not a validated credit model. It remains an input to a regular noncallable bond calculation under the stated payment assumptions.

A sound educational conclusion keeps three statements separate: the size of a hypothetical yield gap, the effect of a specified discount-rate change and the loss distribution under an explicitly assumed default model. None supplies personal investment advice or evidence about a real issuer. The value of the distinction is analytical clarity: a reader can see what has been calculated and which missing assumptions would be necessary for a different question.

Primary references

1. BIS: Interpreting Sovereign Spreads
2. SEC Investor.gov: Bonds FAQs and Risks