Dollar Convexity and Second-Order Hedging

Estimate how far a car travels using only its current speed, and you'll be wrong the moment it accelerates or brakes — speed alone assumes a straight line through time when the real path curves. Duration does the same thing for a bond: it estimates the price change for a small move in yield as if the price-yield relationship were a straight line at that point. It is a good local estimate, but bond prices actually trace a curve, and convexity is the correction term that captures how much that curve bends — how much the slope itself (duration) changes as yields move.

Dollar convexity is simply convexity expressed in dollars rather than as a percentage, matching the way traders already think in DV01 (dollar value of a one-basis-point move) rather than duration. A trading desk managing a large book cares about dollar figures they can net against other positions, not abstract percentages.

The second-order approximation

$$\Delta P \approx -\text{DV01} \times \Delta y \times 100 + \frac{1}{2} \times \text{Dollar Convexity} \times (\Delta y)^2 \times 10{,}000$$

Here $\Delta P$ is the price change in dollars, $\text{DV01}$ is the dollar price change for a one-basis-point yield move (the first-order, duration-based term), $\Delta y$ is the yield change in decimal (e.g., 0.01 for 100bp), and Dollar Convexity is the second-order term capturing curvature, scaled so the units work out in dollars. In words: the first term is the straight-line estimate from duration alone; the second term is a correction that is always positive for an ordinary bond (since convexity is positive), meaning duration alone understates the price gain on a rally and overstates the price loss on a sell-off — convexity is a benefit the investor gets for free with an ordinary, option-free bond.

Worked example: hedging with duration alone falls short

A trader holds $100 million face of a bond with DV01 of $85,000 per basis point and dollar convexity of $40 (in the appropriate scaled units for the formula above). Yields fall 100 basis points ($\Delta y = -0.01$).

Duration-only estimate: DV01 is already the dollar move per 1 basis point, so for a 100 basis point move the first-order price change is simply $85{,}000 \times 100 = 8{,}500{,}000$, i.e. $8,500,000.

Convexity correction: $\frac{1}{2} \times 40 \times (100)^2 = \frac{1}{2} \times 40 \times 10{,}000 = 200{,}000$, adding $200,000 to the estimate.

Total estimated gain: $8,700,000, versus $8,500,000 from duration alone — the convexity term added roughly 2.4 percent to the accuracy of the estimate, and that gap grows much larger for bigger yield swings.

Worked example: comparing two bonds with equal duration

Bond A and Bond B both have DV01 of $50,000, but Bond A has higher convexity (a bullet Treasury) while Bond B has lower, even negative, convexity (a callable agency bond, whose issuer can redeem it early when rates fall, capping the price gain). For a 200 basis point rally, duration alone predicts identical $10 million gains for both. Once convexity is included, Bond A might gain an extra $800,000 beyond the duration estimate, while Bond B, with its embedded call working against the holder, might gain only an extra $100,000 or even see the estimate reduced. Two bonds with "the same interest-rate risk" by DV01 alone behave very differently once the market actually moves.

What this means in practice

Trading desks manage books to be duration-neutral (matching DV01 across long and short positions) but also track convexity separately, because a duration-neutral portfolio that is short convexity — common when a book is long callable bonds or MBS and short Treasuries to hedge — will bleed value in large moves in either direction, gaining less on rallies and losing more on sell-offs than the DV01 hedge alone suggests. Convexity is bought and sold explicitly in the swaption and Treasury options markets for exactly this reason.

Key terms

• DV01 — the dollar price change of a position for a one-basis-point move in yield; the first-order, duration-based sensitivity.
• Dollar convexity — the second-order term capturing how DV01 itself changes as yields move, expressed in dollars.
• Positive convexity — the curve bends in the investor's favor (larger gains on rallies than losses on sell-offs); typical of option-free bonds.
• Negative convexity — the curve bends against the investor, typical of callable bonds and MBS with embedded prepayment options.