Duration and Interest-Rate Risk: The One Number That Predicts the Damage

It is the most useful single number in fixed income, and it is routinely confused with maturity — which it resembles, is measured in years like, and is not. The pricing article established that bond prices move inversely to yields; duration quantifies how much. This article covers what duration actually measures, the three variants a reader will encounter and when each applies, the rule of thumb that makes it immediately usable, and the several honest limitations that stop it being a complete answer. As with everything in this cluster: this is arithmetic describing sensitivity, not a view on where rates are going, and nothing here suggests any reader should hold more or less of anything.

What duration measures, and why it is not maturity

The intuition first. A bond is a set of payments spread over time, and its price is their present value. When the discount rate changes, payments further in the future are affected more than nearer ones — so a bond's price sensitivity depends on where its money sits in time, weighted by how much of the value each payment represents. That weighted average timing is Macaulay duration, measured in years: it is the average time, value-weighted, until the bondholder receives their money. And it is shorter than maturity for any coupon-paying bond, because coupons deliver value before the final date — which is precisely why maturity is the wrong number. Two bonds maturing in ten years, one paying a 6% coupon and one paying 2%, have materially different durations, because the high-coupon bond returns more of its value early. Zero-coupon bonds are the limiting case: with a single payment at the end, duration equals maturity exactly, which is why they are the most rate-sensitive instruments of any given term. The relationships that follow are worth internalising because they explain most of what bonds do. Longer maturity means longer duration (more distant payments). Lower coupon means longer duration (less value returned early). And higher yields mean shorter duration — a subtler point: when the discount rate is high, distant payments contribute proportionally less to present value, so the weighted average timing shortens. Hence a portfolio of long, low-coupon bonds in a low-yield environment carries the maximum possible sensitivity, which is a mechanical statement rather than a warning, and it is why the bond losses of rising-rate periods have been largest exactly where those three conditions coincided.

The three variants, and the rule of thumb

Macaulay duration is the weighted average time to receipt, in years — conceptually clean, and not directly a price-sensitivity figure. Modified duration is the one that answers the practical question: it is Macaulay duration adjusted for the yield, and it estimates the percentage price change for a one-percentage-point change in yield. This gives the rule of thumb that makes duration immediately usable: a bond with a modified duration of 7 will fall roughly 7% in price if yields rise by one percentage point, and rise roughly 7% if they fall by one. For smaller moves, scale proportionally — a 25 basis point move implies roughly 1.75%. That single sentence is most of what a non-professional needs from this subject. Effective duration is the third variant, required whenever a bond has embedded options: because a callable bond's cash flows depend on whether it is called, its sensitivity cannot be computed from a fixed schedule, and effective duration measures the actual modelled price response instead. For callable bonds, high-yield bonds, and mortgage-related securities, effective duration is the only meaningful figure, and modified duration overstates the upside. Two further notes. Duration is additive across a portfolio, weighted by value — which is why funds and portfolios quote a single duration figure, and why that figure summarises rate exposure in one number. And duration is a linear approximation of a curved relationship: the convexity the pricing article introduced is the correction term, and it means duration slightly understates price gains when yields fall and slightly overstates losses when they rise. For small moves the approximation is excellent; for large ones it drifts, and professionals use both figures together.

The limitations — four of them, and they matter

First, duration assumes a parallel shift in the whole yield curve. It answers "what if all rates move one point?" — but real curves twist, steepen, and flatten, with short and long rates moving by different amounts or in opposite directions, which is the yield-curve article's subject. A single duration figure cannot capture a curve reshaping, and professionals decompose exposure by maturity segment (key-rate durations) for exactly this reason. Second, duration says nothing about credit. It measures sensitivity to interest rates only; a bond can be destroyed by a spread widening or a default while its duration figure sits there unchanged, and in high-yield markets credit moves typically dominate rate moves. Duration and credit are separate risks requiring separate measures — the same separation the ratings article insisted on from the other direction. Third, duration is a snapshot that changes. It shortens as a bond ages, shifts as yields move, and jumps when a call becomes likely — so a duration figure has an as-of date like any other field in this portal. And fourth — the point most worth making — duration measures price sensitivity, not the outcome a holder experiences. An investor holding a bond to maturity receives the contracted payments regardless of what duration predicted about interim prices; duration describes what happens if you have to sell, or what happens to a portfolio's marked value. Which matters enormously to a fund reporting monthly, and considerably less to someone whose holding period matches the bond's life. That distinction is the reason this article contains no advice: whether a given duration is appropriate depends entirely on when the holder needs the money, which is personal information this portal does not have. The mechanics belong here; the judgment belongs with the reader and, where wanted, a licensed adviser.