Credit Spreads Tighten, Yet Bonds Fall: Separate Rates, Spreads and Default Risk

A corporate bond can become more expensive relative to government debt while losing value in your account. The apparent contradiction comes from confusing the credit spread with the bond’s total yield. A spread is a relative price signal, not a complete return forecast—and certainly not a default probability printed on a screen. To interpret it, separate the benchmark rate, compensation beyond that benchmark, and the contractual cash flows being priced.

Start with a matched comparison, not the largest yield

In a simple yield comparison, a corporate bond yielding 5.50% against a matched government reference of 4.00% has a spread of 1.50 percentage points, or 150 basis points. That is not a 1.50% capital gain and not the difference between the two coupon rates. Yield depends on price as well as promised payments. An old bond with a high coupon can still offer a low yield when purchased at a premium.

Use the same currency, comparable remaining maturity, valuation time and compounding convention. Then check seniority, collateral and embedded options. Comparing a callable subordinated bond with a short government bill does not isolate one company’s credit risk. A stale corporate trade against a fresh benchmark quote can manufacture a spread move even when no new corporate transaction has occurred.

Reprice the bond: tighter can still mean a loss

Consider a hypothetical five-year bond with face value 100, an annual coupon of 5.50 and principal repayment at maturity. It has no call or put option. Value it immediately after a coupon date, with no accrued interest. Hold all payment dates and promised cash flows fixed while changing yields at the same valuation instant; this is a price sensitivity exercise, not a year of investment performance.

The second row narrows the spread by 20 basis points but raises the benchmark by 50. The total yield rises by 30 basis points, and the price falls about 1.27%. In the third row, a wider spread coexists with a price gain because the benchmark declines more than the spread rises. Neither row alone proves that the issuer has become financially stronger or weaker.

The prices are reproducible: discount each of the five 5.50 coupons at the annual yield, then add 100 discounted for five years. At 5.80%, that gives 98.72938547. The calculation prices promised cash flows; it does not guarantee that an issuer will actually make every payment.

Use two risk contributions without counting the move twice

The starting modified duration is approximately 4.2703. A first-order calculation gives a benchmark contribution of −4.2703 × 0.005, or about −2.1351%, and a spread contribution of −4.2703 × (−0.002), or +0.8541%. Together they give −1.2811%, close to the exact −1.2706%. The difference is the curvature omitted by a linear approximation.

Those sensitivities coincide only under this deliberately simple, option-free, flat-curve setup. Real portfolios require separate interest-rate and spread sensitivities, and often maturity-bucket exposures. Do not add the benchmark loss to a duration loss already calculated from the full 30-basis-point yield change: that counts the rate move twice.

For the conversion into cash, see how DV01 turns duration into money per basis point. An interest-rate hedge can reduce benchmark exposure while leaving spread widening, default jumps, funding and imperfect hedge matching. It is not a guarantee against bond losses. Over an actual holding period, add coupon income, accrued-interest changes, curve roll-down, expenses and realized credit events to the price bridge.

Choose the spread measure before interpreting the chart

A simple matched-maturity yield spread is easy to read, but it compresses a whole cash-flow schedule into one yield. A zero-volatility spread instead adds a constant increment to the benchmark spot curve to reproduce the price of the specified payments. An option-adjusted spread uses a model that also accounts for embedded options. These are different measurements, not interchangeable labels.

For a callable bond, changing interest-rate volatility or the exercise model can alter the option-adjusted result. “Adjusted” does not mean free of assumptions. Record the curve, model and price side before comparing two dates. A wider bid–ask spread is a transaction-cost signal; it is not the same object as the yield spread over government debt, although liquidity can influence both.

Why 150 basis points is not a 1.50% default probability

To see the identification problem, use a separate one-year zero-coupon example. It promises 100 at year-end. The risk-free rate is 4%, and its quoted yield is 5.50%, so the price is 100 ÷ 1.055 = 94.7867. Assume default is resolved only at year-end, recovery equals a fraction R of face value, and there are no liquidity, tax or option effects.

If q is the risk-neutral default probability used for pricing, the discounted expected payment is [100 × (1−q) + 100 × R × q] ÷ 1.04. Equating that to the observed price gives q = 0.015 ÷ [1.055 × (1−R)].

The same spread supports different implied probabilities when recovery changes. Higher recovery requires more assumed defaults to justify the same discounted price in this model. The familiar shortcut spread ÷ loss given default gives 2.50% at 40% recovery, rather than this discrete model’s exact 2.37%. Neither number is an estimate of real-world default frequency. Risk-neutral probabilities reflect pricing; real probabilities require a separate estimation framework. Applying the formula to an actual spread that also contains liquidity and risk premiums adds another identification error.

What would confirm—or invalidate—the signal?

First freeze the instrument and quote conventions. Then reconcile the benchmark and spread changes to the bond’s own price using appropriate sensitivities. Check whether multiple executable quotes or recent comparable trades support the move, and distinguish issuer-wide changes from one illiquid security.

For an index, examine constituent changes, rating migrations and duration shifts. Removing distressed issuers can narrow an index spread without improving the surviving bonds’ contractual protection. For one issuer, inspect near-term maturities, operating cash flow, refinancing access, collateral and covenants. A rating label alone is not the analysis.

The interpretation fails when quotes are stale, the benchmark or option model changes, or unlike instruments are compared. If those checks pass, a tighter spread says the market is demanding less compensation relative to the chosen benchmark. Whether that is attractive still depends on the remaining cushion, recovery assumptions and the rate-and-spread stress the portfolio can absorb.