Bond Duration and Convexity Explained (With Examples)

TL;DR. Duration is the number that tells you, in one figure, how much a bond’s price should move when yields move. The rule of thumb: for a small change in yield, price moves by roughly −modified duration × Δy. A 7-year modified duration and a 1-percentage-point rise in yield implies about a −7% price move. Convexity is the correction term — a second derivative that captures the fact that the price–yield curve is bowed, not straight, so duration alone always underestimates price when yields fall and overestimates price when yields rise. Long-dated zero-coupon bonds have the highest duration; short T-bills the lowest. That is why long Treasuries lost roughly a third of their value in 2022 while T-bill investors barely noticed the same rate cycle.

Why duration exists at all

Bond prices move inversely to yields — that much every investor learns early. What is less obvious is how much a bond’s price should move for a given change in yield. The answer depends on the bond’s coupon, maturity, yield level, and payment frequency, all mashed together. Duration is the industry’s single-number summary of that sensitivity.

Frederick Macaulay introduced the concept in 1938 as the present-value-weighted average time to a bond’s cash flows. Modern practice uses a related figure, modified duration, that is directly interpretable as a percentage price move per unit yield move. Both are widely defined in fixed-income textbooks and in public reference materials such as the Wikipedia bond-duration entry.

The U.S. Securities and Exchange Commission’s Office of Investor Education has warned since at least 2013 that duration is the primary tool for gauging a bond portfolio’s interest-rate risk and that longer-duration portfolios can suffer “significant” losses when rates rise (SEC Investor Bulletin — Interest Rate Risk).

The three durations worth knowing

Practitioners use three related quantities. All three have units of years, and all three fall as coupons rise or as maturity shortens.

Macaulay duration

The original definition — the present-value-weighted average time to a bond’s cash flows:

DMac = ∑i ( ti · PV(Ci) ) ÷ Price

Each cash flow Ci arriving at time ti contributes its present value as a weight. A zero-coupon bond has the maximum possible Macaulay duration for its maturity: since the only cash flow is the face value at maturity, its duration equals its maturity.

Modified duration

Macaulay duration is a weighted average time; modified duration is what most desks actually quote. It is Macaulay duration divided by (1 + y/m), where y is the yield to maturity and m is the compounding frequency:

Dmod = DMac ÷ (1 + y/m)

The reason it matters is that modified duration is the first derivative of price with respect to yield, divided by price, with a sign flip. Which gives the first-order price-change rule of thumb everyone quotes:

ΔPrice/Price ≈ −Dmod × Δy

So a bond with modified duration 7.0 should lose about 7% of its price if yields rise by 100 basis points, and gain about 7% if yields fall by 100 basis points. Note that this is only a first-order approximation — and it gets worse as Δy gets bigger. That is where convexity comes in.

Effective duration

For bonds with embedded options — callable corporates, mortgage-backed securities, prepayable munis — Macaulay and modified duration mislead because the cash-flow schedule changes as yields change. Effective duration measures sensitivity numerically by shocking the yield curve up and down by a small amount and observing what a full valuation model does to the price:

Deff ≈ ( PP+ ) ÷ ( 2 · P0 · Δy )

where P and P+ are model prices after small down- and up-shifts. For plain-vanilla non-callable bonds, effective duration collapses to modified duration; for a mortgage-backed pool it can be dramatically different (bond duration — effective duration).

Convexity: the correction term the straight-line rule misses

The price–yield relationship of a normal bond is a smooth downward-sloping curve, not a straight line. Modified duration draws the tangent to that curve at today’s yield. For very small yield moves the tangent is fine. For larger moves the curve bows away from the tangent — and always in the investor’s favour: prices fall by less than duration predicts when yields spike, and rise by more than duration predicts when yields collapse. This bowing is called convexity, and it is defined as the second derivative of price with respect to yield, normalised by price:

C = ( 1 ÷ Price ) · d2Price ÷ dy2

The full second-order approximation to a bond’s price change is:

ΔPrice/Price ≈ −Dmod · Δy + ½ · C · (Δy)2

Two things to notice. First, convexity is always positive for a plain-vanilla bond, so the correction term always adds to the price estimate. Second, it scales with (Δy)2, which is negligible for a 5-basis-point move and material for a 200-basis-point move. Long-dated bonds have far more convexity than short-dated bonds because their price–yield curves are far more bowed.

A worked example on a 10-year Treasury

The Federal Reserve H.15 release for the week ending August 11, 2026 shows the 10-year U.S. Treasury constant-maturity yield at 4.70% (Federal Reserve H.15). Take a stylised 10-year U.S. Treasury note with a 4.70% coupon, priced at par ($100) and paying semiannually, so y = 4.70% and m = 2.

Working through the standard bond-math for this par bond gives:

  • Macaulay duration DMac8.06 years
  • Modified duration Dmod = 8.06 ÷ (1 + 0.047/2) ≈ 7.87 years
  • Convexity C77 (using the standard bond-math convention)

Now suppose yields rise by 100 basis points, from 4.70% to 5.70%. The duration-only estimate predicts a price change of:

−7.87 × 0.01 = −7.87%

The full second-order estimate adds the convexity correction:

−7.87 × 0.01 + ½ × 77 × (0.01)2 = −7.87% + 0.385% = −7.49%

The actual, exactly-repriced move on a 4.70%-coupon 10-year par bond going to a 5.70% yield is a price of about $92.44, a loss of −7.56%. The straight-duration rule over-predicts the loss by 31 basis points; the duration + convexity estimate under-predicts it by only 7. For a 200-basis-point shock the convexity term is roughly four times as large (it scales with the square) and matters a great deal more.

Duration rises with maturity, falls with coupon

Two intuitions the formula makes rigorous:

  • Longer maturity ⇒ higher duration. Cash flows further in the future carry more weight and are discounted more heavily, so distant maturity dominates the calculation. A 30-year par bond has a duration close to 15 years; a 2-year par bond has a duration close to 2.
  • Higher coupon ⇒ lower duration. A larger coupon returns more of your money sooner, shifting the weighted average time forward and shortening duration. A zero-coupon bond gets no coupons at all, so its Macaulay duration equals its maturity — the highest possible for its tenor.

The table below shows illustrative modified durations for the current Treasury yield curve, using each tenor’s constant-maturity yield from the latest H.15 release as both the coupon and the discount rate (a par-bond assumption).

Tenor Yield (%) Modified duration (yrs) Est. price move on +100 bps
3-month T-bill 3.89 0.25 −0.25%
2-year note 4.22 1.90 −1.90%
5-year note 4.39 4.42 −4.42%
7-year note 4.54 5.94 −5.94%
10-year note 4.70 7.87 −7.87%
20-year bond 5.25 12.35 −12.35%
30-year bond 5.24 15.42 −15.42%
Yields: Federal Reserve H.15 selected interest rates, week ending August 11, 2026. Durations are par-bond, semiannual-coupon approximations for illustration only; on-the-run Treasury durations differ modestly from these figures.

The price–yield curve is bowed — that’s convexity in one picture

Price versus yield for a 10-year par bond, with the duration tangent line The exact price-yield curve is bowed (convex); the duration approximation is a straight tangent line drawn at the current yield. Duration alone over-predicts losses on the way up and under-predicts gains on the way down.

130 120 110 100 90

2.7% 3.4% 4.0% 4.7% 5.4% 6.0%

Par: yield 4.70%, price $100

Exact price–yield curve Duration tangent (straight-line)

Yield to maturity Price ($)

Illustration for a 10-year 4.70%-coupon par bond. The tangent line is what modified duration alone predicts; the gap between the tangent and the true curve is convexity. Curve values computed from the standard semiannual bond-pricing formula.

The two curves touch only at today’s yield (4.70% in this example). Move either way and the tangent line falls below the true curve — meaning the straight-line duration estimate under-predicts the price for any move in either direction. That gap is small for small moves and grows with the square of the yield change.

What history looks like when duration is very long

Bloomberg U.S. Aggregate Bond Index annual total returns, 2019–2023 2022 was the Bloomberg U.S. Aggregate’s worst calendar year on record, driven almost entirely by the sharp back-up in yields interacting with a portfolio duration of roughly six years.

+10% +5% 0% −5% −13%

+8.72% 2019

+7.51% 2020

−1.54% 2021

−13.01% 2022

+5.53% 2023

Bloomberg U.S. Aggregate Bond Index — annual total return

Data: Bloomberg U.S. Aggregate Bond Index calendar-year total returns as reported in Federal Reserve academic and industry commentary. The index carries a duration of roughly six years; 2022’s ~200-basis-point back-up in yields drove the worst calendar year in the index’s history (Federal Reserve FEDS Notes, August 2024).

2022 is the cleanest recent object lesson in duration. The Bloomberg U.S. Aggregate Bond Index, which most balanced funds use as a fixed-income benchmark, has a duration of roughly six years. Its 2022 loss of about 13% is very close to what its duration times the year’s ~200-basis-point yield back-up would predict — with a small offset from convexity and the coupon return in the other direction.

Common mistakes retail investors make with duration

  • Confusing duration with maturity. They are not the same. A 30-year bond with a 5% coupon has a duration around 15 years, not 30. A 30-year zero-coupon bond’s Macaulay duration is 30 years — and its price move for a 100-basis-point yield change is proportionally larger.
  • Assuming the duration rule holds for big yield moves. For a 25-basis-point shock, ignoring convexity costs almost nothing. For a 200-basis-point shock in a long-duration bond, ignoring convexity can misprice the position by several percent.
  • Using modified duration on callable or MBS securities. Cash flows change with rates in these securities. Use effective duration (or a full valuation model). Mortgage-backed securities can even have negative convexity at low yields because falling rates trigger refinancing.
  • Forgetting duration changes as time passes. A bond’s duration shortens every day as its cash flows draw closer. Duration also changes with yield: as yields rise, duration falls modestly (the “convexity of duration”).
  • Ignoring the coupon reinvestment side. Duration captures the price-risk half of interest-rate risk. Higher yields hurt price today but lift future coupon reinvestment income — the two effects offset over a holding period equal to the bond’s Macaulay duration. This is the classic duration immunization result.

Where duration sits in the bigger picture

Duration is the fixed-income analogue of beta on the equity side: a one-number sensitivity to the market’s dominant risk factor. Portfolio managers extend the idea in two natural directions. Key-rate durations (or partial durations) break out sensitivity to shifts at specific points on the yield curve — useful for portfolios that face non-parallel curve moves such as steepeners and flatteners. Spread duration captures sensitivity to credit-spread moves (as opposed to Treasury yields) and matters most for corporate and high-yield portfolios where credit spreads dominate the risk budget.

Related concepts and what to learn next

Sources

Disclosure: This article is for informational purposes only and is not investment advice.

Leave a Comment