Credit spreads

~9 min read

🎯 Interview frequent

Why it matters

A corporate bond’s yield is the risk-free rate plus a spread that compensates for everything a Treasury doesn’t have to worry about: default risk, liquidity, and structural complexity. “Spread” is the single number a credit desk actually trades and monitors intraday — decomposing it correctly is a recurring interview centerpiece.

Core concepts

The spread decomposition. Corporate yield \(\approx\) risk-free rate + credit spread. But the credit spread itself is not pure expected loss — it bundles expected loss (default probability × loss given default), a risk premium for bearing that uncertain loss, and liquidity/technical compensation (corporates trade far less than Treasuries). Risk-neutral (market-implied) default probabilities extracted from spreads are almost always higher than real-world historical default rates for the same rating — the same wedge as implied vs. realized volatility in options.
Nominal spread is simply YTM(corporate) − YTM(comparable-maturity Treasury). It’s easy to compute but ignores the shape of the yield curve — a poor tool when the curve isn’t flat.
Z-spread (zero-volatility spread) is the constant spread added to every point on the Treasury spot curve that makes the discounted cash flows equal the bond’s price. It’s curve-consistent, unlike the nominal spread, and is the standard credit-desk convention for a bullet bond.
OAS (option-adjusted spread) strips out the value of embedded optionality (calls, puts) from the Z-spread, using an interest-rate model to value the option first. OAS is the apples-to-apples spread for comparing a callable corporate to a bullet — Z-spread alone conflates call risk with credit risk.
Asset swap spread reframes the bond versus the swap curve rather than Treasuries — closer to a bank or hedge fund’s actual funding curve, and common in comparing corporates to floating-rate benchmarks.
CDS-bond basis. Buying a cash bond and buying CDS protection on the same name is roughly a synthetic risk-free position (plus funding). Basis = CDS spread − bond’s Z-spread (sign convention varies by desk). A persistently nonzero basis reflects funding costs, cheapest-to-deliver optionality in the CDS, counterparty risk, and technical supply/demand — not just credit.
Spread widening/tightening — desk language. “The name widened 20bp” means the spread increased — bad for anyone long the bond or short protection, since the market now demands more compensation for the same credit. Spread moves can happen with zero change in the risk-free curve, which is exactly why spread duration is tracked separately from rate duration.

Mental model

  Corporate yield = Risk-free rate  +  Credit spread
                                          |
                     -----------------------------------------
                     |               |                       |
              Expected loss    Risk premium          Liquidity / technical
              (PD x LGD)      (compensation for      (corporates trade
                               uncertainty of loss)   thinner than UST)

  Nominal spread:  one number vs. one Treasury point   (curve-blind)
  Z-spread:        one number added to the WHOLE curve  (curve-consistent)
  OAS:             Z-spread minus the value of embedded optionality

Think of spread the way you think of an insurance premium: part of it is the actuarially fair cost of the loss, and part of it is what the insurer charges you for being the one stuck holding an uncertain claim.

Interview questions

A corporate bond’s Z-spread is 150bp. What does that number actually represent? Answer: The constant spread that, added uniformly to every point on the Treasury spot curve, makes the discounted contractual cash flows equal the bond’s market price. It’s a curve-consistent measure of compensation over risk-free, bundling default risk, liquidity, and any residual technical factors.
Why would a bond’s Z-spread and OAS differ meaningfully, and for which bonds should they be closest? Answer: They differ when the bond has embedded optionality (a call, typically) — OAS removes the value of that option from the spread, so OAS < Z-spread for a callable bond when the call has positive value to the issuer. For a non-callable bullet, there’s no option to strip out, so Z-spread and OAS should be nearly identical.
Risk-neutral default probabilities implied from CDS/bond spreads are usually higher than historical (real-world) default rates for the same rating. Why? Answer: Spreads price under a risk-neutral measure that embeds a risk premium (investors demand extra compensation for bearing correlated, fat-tailed default risk) plus liquidity and dealer balance-sheet costs. Historical default rates are simply realized frequencies under the real-world measure — the same “Q vs. P” wedge that shows up between implied and realized volatility.
A trader tells you “the CDS-bond basis on this name is very negative.” What’s actually going on, and what trade does that suggest? Answer: CDS spread is trading tight relative to the bond’s Z-spread (protection is “cheap” relative to the cash bond’s implied credit risk). It can reflect funding costs on the bond, cheapest-to-deliver value in the CDS, or technical supply/demand. A basis trade would buy the cash bond and buy protection to capture the gap, subject to funding and counterparty risk.
Why do desks track spread duration separately from interest-rate duration for a corporate bond? Answer: A bond’s spread can move with zero change in the Treasury curve — the market can reprice the same issuer’s credit risk independently of rates. Spread duration measures price sensitivity to that spread move specifically, and for a stressed or long-dated credit it can dominate rate duration as a source of P&L risk (see Rate risk vs. credit risk).

Watch

Session 11: Costs of Debt and Capital — Prof. Aswath Damodaran (NYU Stern). Walks through debt ratings and synthetic default spreads as the building blocks of a corporate’s cost of debt.
Credit default swaps — Khan Academy. The running spread as the price of protection, and how it packages default compensation.