Convexity of Convexity

What Is Convexity of Convexity?

Convexity of Convexity — sometimes called second-order convexity or informally the "gamma of duration" — is the rate of change of a bond's convexity with respect to a shift in yield. While standard convexity measures how a bond's duration changes as yields move (the second derivative of price with respect to yield), convexity of convexity captures the third derivative: how that curvature itself accelerates or decelerates under large yield shocks.

In formal terms, if price P is a function of yield y, standard duration is -dP/dy, convexity is d²P/dy², and convexity of convexity is d³P/dy³ — the third-order Taylor expansion term, scaled appropriately. For vanilla bullet bonds, this term is often negligible. But for mortgage-backed securities, callable bonds, structured products, and options-embedded instruments, it becomes measurable and at times tradeable in its own right.

Why It Matters for Traders

For most investors, standard duration and convexity are sufficient to manage rate risk under moderate yield moves of 25–75 basis points. However, when yields shift violently — as they did in 2022, when 10-year Treasury yields moved from roughly 1.5% in January to over 4.2% by October — the third-order term becomes non-trivial in portfolios holding negative convexity instruments.

Mortgage servicers and bank asset-liability managers who rely purely on convexity-adjusted duration hedges find that their hedge ratios drift faster than models predict under sustained rate moves. This is precisely because they are not accounting for the convexity of convexity — their hedges must be rebalanced not just when yields move, but at an accelerating rate relative to what first and second derivatives alone would imply. Convexity hedging desks at major primary dealers implicitly trade this risk when rolling large MBS duration hedges.

How to Read and Interpret It

Convexity of convexity does not have a universal quoted metric the way duration or convexity does — it is computed analytically from a bond's cash flow structure and embedded option model. Key interpretive rules:

• A positive third-order term means a bond's convexity is increasing as yields rise — a favorable self-reinforcing property that benefits long-only holders.
• A negative third-order term, common in agency MBS due to prepayment optionality, means convexity erodes faster than expected at the extremes, dramatically underperforming a naive hedge.
• For practical desk risk, thresholds of concern emerge when yield moves exceed ±150 bps within a quarter, at which point even well-hedged MBS books can show unexplained P&L gaps attributable to this higher-order effect.
• Structured credit desks often stress-test third-derivative sensitivity separately in their Greeks reporting under +300 bps and -200 bps parallel shift scenarios.

Historical Context

The 1994 bond market selloff provides the clearest historical case. The Federal Reserve raised rates by 300 basis points between February 1994 and February 1995, catching mortgage portfolios catastrophically under-hedged. Duration and convexity models of the era predicted manageable losses, but the convexity of convexity effect in MBS meant that prepayment speeds and option-adjusted durations shifted non-linearly. Orange County's investment pool, which lost approximately $1.7 billion, held structured notes whose convexity of convexity was deeply negative and essentially unmodeled at the time — the portfolio's sensitivity accelerated in precisely the wrong direction as rates rose.

The 2022 rate shock similarly exposed regional bank held-to-maturity portfolios whose embedded convexity of convexity had not been stress-tested under 400 bps moves, contributing to the conditions that led to Silicon Valley Bank's collapse in March 2023.

Limitations and Caveats

This metric is model-dependent: the third derivative is sensitive to the prepayment or option model used, and small changes in model assumptions can dramatically alter its sign and magnitude. For sovereign bullet bonds with no embedded options, convexity of convexity is near zero and effectively irrelevant. Furthermore, computing it reliably requires granular scenario analysis rather than closed-form solutions for most structured instruments, making real-time risk monitoring challenging.

What to Watch

• Federal Reserve pace of rate changes: the faster and larger the move, the more this term matters.
• Agency MBS prepayment model revisions from GSEs, which directly alter the implied convexity surface.
• Bank ALM disclosures for signals that held-to-maturity portfolio hedges are drifting relative to stated duration targets, which may indicate unmodeled higher-order risk.
• Dealer convexity hedging flow activity in swaption markets, particularly 1y10y and 3m10y vols.