Guide

Option-adjusted spread (OAS) explained

Harbor Capital’s fixed-income desk held two agency MBS pass-throughs trading at nearly identical dollar prices: a FNMA 5.5% 30-year pool at 98.25 and a FNMA 6.0% pool at 98.40. Headline credit spreads over Treasuries looked comparable — 145 bps vs 152 bps — but the 5.5% coupon carried far more negative convexity: when the 10-year Treasury yield fell 40 bps in two weeks, the 5.5% pool lagged the 6.0% by 1.8 points of total return. The risk committee asked a sharper question: after stripping out the prepayment option homeowners hold, which pool offered better compensation per unit of rate risk? The answer required option-adjusted spread (OAS), not raw yield spread.

OAS is the constant spread added to the risk-free forward curve in a valuation model such that the model price equals the market price — after explicitly modeling embedded options (issuer calls, investor puts, prepayment, conversion). Unlike a simple Z-spread that assumes cash flows are fixed, OAS answers: “How much extra yield am I earning for credit, liquidity, and complexity once I have paid fair value for the options?” This guide defines OAS vs Z-spread and asset-swap spread, walks through binomial and Monte Carlo trees for callable corporates and MBS, covers effective duration and convexity off the OAS curve, the Harbor Capital agency sleeve refactor, a technique decision table, pitfalls, and a portfolio checklist.

Why raw yield spread misleads

A corporate bond callable in two years at par looks cheap on yield-to-worst versus a non-callable sibling from the same issuer. The callable trades wider because investors know the issuer will call if rates fall — capping upside. Comparing the two on nominal spread to Treasuries mixes credit risk with option value.

Three spread measures appear on Bloomberg and in fund fact sheets:

G-spread / nominal spread — yield difference vs a constant-maturity Treasury. Fast but ignores curve shape and options.
Z-spread (zero-volatility spread) — flat spread over the entire swap or Treasury spot curve that discounts fixed projected cash flows. Works for bullets; breaks when prepayments or calls change paths.
OAS — spread after modeling stochastic rates and optimal exercise of options along each path.

For plain-vanilla Treasuries and non-callable bullets, Z-spread and OAS converge. For agency MBS, CLO tranches with prepayment, and callable corporates, the gap between Z-spread and OAS can exceed 30–50 bps — enough to invert relative-value rankings.

How OAS is computed

The tree intuition

Practitioners build a short-rate tree or use Monte Carlo paths calibrated to the current swap curve and implied cap/floor volatilities. At each node the model prices the bond by rolling back expected cash flows, applying call or prepayment rules when exercise is optimal for the option holder.

OAS is solved iteratively: add a trial spread to every risk-free rate on the tree, reprice the bond, adjust the spread until model price equals market price. The spread is “option-adjusted” because the tree already valued the embedded option; OAS captures everything else — prepayment model error aside, it is the cleanest read on compensation for holding complex cash flows.

Callable corporate example

A 10-year BBB corporate with a 5.75% coupon callable at par in year 3 trades at 99.50. Z-spread to the swap curve might print 185 bps. On a Hull-White or Black-Derman-Toy tree with issuer call at par when rates drop, the model fair value at zero OAS might be 101.20 — the call caps price appreciation. Backing out OAS yields 142 bps: the bond is tighter on an option-adjusted basis than Z-spread suggests because a large chunk of the nominal spread compensates for the call, not credit.

MBS and prepayment models

Agency MBS OAS depends critically on the prepayment model (PSA ramps, turnover, refinancing S-curves tied to incentive spread). Two desks can quote different OAS on the same TBA pool if their CPR assumptions diverge. That is why OAS is a relative-value tool within a consistent model, not an absolute truth stamped on the security.

Effective duration and convexity off OAS

Modified duration assumes parallel yield shifts on fixed cash flows. Callable bonds and MBS exhibit negative convexity: prices rise less than duration predicts when rates fall (calls/prepays accelerate) and can fall faster when rates rise (extension risk).

Effective duration bumps the OAS curve up and down by a small amount (often 25 bps), re-runs the tree, and measures percentage price change. Effective convexity captures the curvature. A FNMA 5.5% 30-year pool might show effective duration of 5.8 years and effective convexity of −1.2 at current OAS — very different from a 10-year Treasury with positive convexity near +0.8.

Portfolio managers sum effective duration across MBS and corporates to hedge rate risk with Treasury futures or swaps. Ignoring effective metrics and using static duration on callable sleeves systematically misstates hedge ratios.

Harbor Capital agency MBS sleeve refactor

Harbor’s $240M agency MBS sleeve was overweight 5.0% and 5.5% coupons bought on nominal spread cheapness after the 2023 rate spike. Risk reported effective duration of 4.9 years, but OAS relative-value screens showed those coupons offered only 8–12 bps OAS advantage over 6.0% pools despite carrying 0.4 years more effective duration — poor risk-adjusted carry once prepayment option cost was netted out.

The refactor:

Standardized on a single vendor prepayment model and monthly OAS grid for all agency holdings.
Ranked pools by OAS per unit effective duration rather than nominal spread to Treasuries.
Rotated $38M from low-coupon pools into 6.0% and 6.5% TBAs with higher OAS and shorter effective duration at the prevailing curve.
Hedged 60% of effective duration with 5-year Treasury futures, sized off OAS-based effective metrics, not static WAL duration.

Over the next quarter, a 55 bps rally in the 10-year Treasury produced +2.1% total return on the refactored sleeve vs +0.6% on a static low-coupon proxy portfolio. The lesson: OAS ranking prevented owning bonds that looked wide but were mostly paying for negative convexity the desk did not want.

Technique decision table

Metric	Use when	Skip when
OAS (tree/Monte Carlo)	Callable corporates, MBS, ABS with prepay, convertible bonds, relative value within optioned sectors	Plain bullets with no options; need sub-second screening of thousands of IG names
Z-spread	Non-callable corporates, fast desk screens, comparing bullets on the same curve	Any security where cash flows shift with rates (MBS, calls, sinks)
Asset-swap spread (ASW)	Comparing bond richness vs funding in swap space; derivatives desks hedging credit	Illiquid bonds with wide bid-ask; securities with complex repo financing
I-spread (swap spread)	Quick IG bond vs swap curve richness without building a tree	Callable or prepay structures; when swap curve basis distorts cross-market reads
CDS-implied spread	Single-name credit hedging, capital-structure arb vs cash bond	Agency MBS with no credit risk; basis risk between CDS and cash is wide

Common pitfalls

Comparing OAS across different prepayment models — a 20 bps OAS gap between desks may be model noise, not opportunity.
Treating OAS as credit spread for agencies — agency MBS OAS is mostly prepayment and liquidity, not default; do not stack it beside corporate OAS without context.
Ignoring convexity at portfolio level — high OAS low-coupon MBS can crater in rallies; effective convexity belongs on the same dashboard as OAS.
Using stale volatility inputs — OAS shifts when swaption vol moves even if price is unchanged; mark-to-model risk is real.
Confusing OAS with Z-spread on callables — ranking callable corporates on Z-spread systematically favors longer call protection that may never pay.
Overfitting the tree to one price — calibrate to market, but stress CPR +20% and −20% for MBS; OAS stability under stress matters more than a point estimate.
Forgetting transaction costs — a 5 bps OAS pickup on a TBA roll may vanish after pay-up and financing.
Mixing Treasury and swap OAS conventions — specify whether the tree is calibrated to SOFR swaps or Treasuries; the level differs.

Production checklist

Document curve (Treasury vs SOFR swap) and vol surface used for every OAS run.
Standardize prepayment model version across MBS and ABS holdings.
Report Z-spread, OAS, effective duration, and effective convexity together.
Rank positions by OAS per unit effective duration, not nominal spread alone.
Stress OAS at CPR ±20% and parallel rate shifts ±50 bps.
Reconcile OAS-based hedge ratios with futures/swaps monthly.
Flag when Z-spread minus OAS exceeds desk threshold (e.g. 25 bps) for review.
Include financing and pay-up in relative-value trade tickets.
Archive model inputs on trade date for audit and post-mortem.
Re-calibrate vol and CPR assumptions after major macro shocks.

Key takeaways

OAS is the spread over the risk-free curve after valuing embedded options — it isolates compensation for credit, liquidity, and model risk from option cost.
Z-spread overstates richness on callable corporates and MBS because it ignores prepayment and call paths.
Effective duration and convexity off the OAS tree are mandatory for hedging optioned fixed income.
Harbor Capital improved rally-period returns by rotating from low-coupon MBS with poor OAS per unit duration into higher-coupon pools.
OAS is model-dependent — use it for relative value within a consistent framework, not as an absolute cross-asset truth.