Convexity-Adjusted Duration

What Is Convexity-Adjusted Duration?

Convexity-adjusted duration extends standard modified duration — which assumes a linear relationship between bond prices and yield changes — by adding a second-order correction term that captures the actual curvature of the price-yield curve. The mathematical formulation is: ΔP/P ≈ −D·Δy + ½·C·(Δy)², where D is modified duration and C is convexity. For small yield moves, the linear duration approximation is adequate; for larger moves typically beyond 50–100 basis points, the convexity term becomes material and ignoring it introduces significant hedging error.

Bonds with positive convexity — standard bullets and zero-coupon Treasuries — gain more than duration predicts when yields fall and lose less than duration predicts when yields rise, a structurally favorable asymmetry that investors pay a premium to own. Negatively convex bonds — primarily mortgage-backed securities, callable corporates, and certain structured products — behave inversely: they underperform duration predictions in both favorable and adverse yield environments because embedded short optionality erodes returns precisely when directional exposure is largest. This is why option-adjusted duration and effective duration — which reflect the full probability-weighted distribution of cash flows across rate scenarios — often diverge meaningfully from modified duration for instruments with embedded options.

Why It Matters for Traders

For macro traders managing duration risk across large fixed income portfolios, the distinction between modified duration and convexity-adjusted duration is not academic — it determines whether a hedge actually works under stress. During the bear steepener of 2022, when 10-year Treasury yields rose roughly 280 basis points from January through October, portfolios hedged using duration alone significantly underestimated actual losses because the magnitude of the move exposed convexity mismatches that simple DV01 hedging could not contain.

Mortgage-backed securities present the starkest case of convexity in practice. As rates rise, homeowner prepayments slow, extending MBS duration precisely when exposure is most costly — this negative convexity makes duration a moving target rather than a fixed parameter. Agency MBS dealers must continuously re-hedge as rates move, generating systematic delta hedging flows that create observable momentum in Treasury markets. Estimates suggest that for every 25-basis-point move in rates, aggregate MBS convexity hedging can generate $40–80 billion in notional Treasury buying or selling — a force large enough to amplify intraday trends and exacerbate yield momentum.

Conversely, long-dated positively convex instruments like 30-year Treasury STRIPS provide natural hedge acceleration: as yields fall, their duration extends mechanically, generating larger-than-predicted gains. Macro funds running long-volatility fixed income strategies often express views through convexity rather than outright duration, using options on Treasury futures or long positions in high-convexity bonds offset by short lower-convexity instruments.

How to Read and Interpret It

For a standard 10-year Treasury, modified duration runs approximately 8.5–9.0 years and convexity is roughly 0.80–1.00. A 100-basis-point parallel shift upward produces a linear duration-predicted loss of about 8.7% but a convexity-adjusted loss closer to 8.3% — convexity adds approximately 40 basis points of price cushion. Scale this to a $1 billion portfolio and the convexity adjustment equals roughly $4 million on a 100-basis-point move — not trivial.

For a 30-year zero-coupon Treasury, convexity exceeds 9.0, making the second-order adjustment enormous for large rate moves. A 200-basis-point yield increase would be overstated by nearly 180 basis points of predicted price loss if convexity is ignored. For a current-coupon agency MBS, convexity may be negative 2 to negative 5, meaning the bond loses more than duration predicts as rates rise and gains less as rates fall. Portfolios carrying aggregate convexity below −3 face significant hedging complexity and potential for forced duration extension selling into rising markets. Practitioners compare option-adjusted duration to nominal modified duration as a direct measure of embedded optionality value: a callable corporate with OAD of 4.5 years versus modified duration of 7.0 years implies substantial call probability at current rate levels.

Historical Context

The 1994 bond market massacre remains the canonical convexity event in fixed income markets. When the Federal Reserve unexpectedly raised rates 25 basis points in February 1994 — the first tightening since 1989 — it triggered a chain reaction that pushed 10-year Treasury yields from approximately 5.6% in January 1994 to nearly 8.0% by November. MBS holders discovered their negative convexity was catastrophic in practice: as rates rose, prepayment speeds collapsed, automatically extending portfolio duration on instruments already suffering mark-to-market losses. Dealers and thrifts forced to hedge extending duration sold Treasuries and entered receiver swaptions, which pushed yields higher still, triggering additional convexity hedging in a self-reinforcing feedback loop.

Orange County, California's investment pool lost approximately $1.7 billion through leveraged exposure to negatively convex structured products, where convexity unwind compounded directional losses far beyond what modified duration estimates would have suggested. The episode codified industry awareness that ignoring second-order terms in large-rate-move environments is not a rounding error — it is a risk management failure.

More recently, the 2013 Taper Tantrum produced a similar but less severe dynamic: 10-year yields rose roughly 130 basis points from May to September, forcing MBS convexity hedging flows that several primary dealers estimated extended the yield move by 20–30 basis points beyond fundamental drivers.

Limitations and Caveats

Convexity adjustment assumes parallel yield curve shifts, but real-world moves involve twists, butterflies, and idiosyncratic spread changes that a single scalar duration measure cannot capture. A steepening move that leaves the 10-year unchanged but raises the 30-year 50 basis points affects a long-duration portfolio very differently than a parallel shift of identical DV01 magnitude.

Option-adjusted convexity for complex instruments — CMOs, callable corporates, convertible bonds — depends heavily on the chosen interest rate model (Hull-White, Black-Karasinski, or two-factor variants), and model calibration error translates directly into hedging error. For very large yield moves beyond 200–300 basis points, even second-order convexity adjustments prove insufficient and full scenario repricing or Monte Carlo simulation becomes necessary. In derivatives portfolios, vanna and volga effects introduce cross-Greeks that interact with convexity in ways the standard duration framework ignores entirely, particularly when managing gamma-heavy options books.

What to Watch

• Aggregate MBS negative convexity as estimated by dealer research — when market-wide MBS convexity falls below −3, systematic hedging flows become a primary driver of intraday Treasury momentum
• The convexity-to-duration ratio in long Treasury positions: a higher ratio signals more asymmetric payoff and better convexity per unit of rate risk
• Fed balance sheet policy on MBS reinvestment — decisions to let MBS roll off versus reinvest alter the total supply of negative convexity held by the official sector versus the market
• Mortgage rate lock pipelines and refinancing applications, which lead MBS prepayment speeds by 30–60 days and signal impending duration and convexity shifts
• Swaption implied volatility: since convexity is mathematically equivalent to long optionality, rising swaption vol increases the theoretical value of convexity and affects relative pricing of bullets versus callables