Why Measure Spread Duration?
A fixed-income asset can lose value even when benchmark interest rates do not change. If investors demand a larger credit spread for bearing the asset’s credit risk, its contractual cash flows are discounted at higher yields and its market value generally falls. Spread duration measures this price sensitivity to a change in credit spread, separating that sensitivity from a change in the benchmark yield curve.
The separation is important for risk measurement: the benchmark curve and the credit spread are distinct pricing inputs. This tutorial holds the benchmark curve fixed and changes only the spread. A spread-duration estimate is a local, first-order estimate; it is not a substitute for full repricing when the shock is large or the instrument has material embedded options.
Cash-Flow Model and Formula
First specify how the spread enters the price. Consider an instrument with deterministic cash flows \(C_t\) paid at times \(t=1,\ldots,n\). Assume annual payments, annual effective compounding, and a flat benchmark yield \(y\). The credit spread \(s\) is added to that yield at every payment date. There are no mortality contingencies, defaults, embedded options, or uncertain cash flows in this model; the \(C_t\) values are fixed. The price is
Here \(y\) and \(s\) are expressed as annual rates in decimal form. For example, 100 basis points is \(0.0100\). This flat-yield formula is a teaching model; with a non-flat benchmark curve, each cash flow is discounted using its corresponding benchmark spot rate plus the specified spread.
Differentiate each discounted cash flow with respect to \(s\). Since the derivative of \((1+y+s)^{-t}\) is \(-t(1+y+s)^{-t-1}\), the price sensitivity and spread duration are
To interpret the result, let \(PV_t=C_t/(1+y+s)^t\) be the present value of the cash flow at time \(t\). Then
Each cash flow’s weight is its share of the instrument’s price, and its sensitivity per unit of weight is \(t/(1+y+s)\). Thus, under this model, spread duration is a present-value-weighted average of modified cash-flow times. Later cash flows generally contribute more sensitivity, although their price weights depend on the discounting and the cash-flow pattern.
Using Duration to Estimate a Price Change
For a small spread change \(\Delta s\), the first-order approximation follows from the definition of the derivative:
The sign matters. A positive \(\Delta s\) indicates spread widening and normally gives a negative estimated price change when spread duration is positive. A negative \(\Delta s\) indicates spread tightening and normally gives a positive estimated price change. Use decimal changes: a widening of 35 basis points is \(\Delta s=0.0035\), not \(0.35\) or \(35\).
Worked Example: Calculate Duration and Estimate a Bond’s Spread Loss
A hypothetical two-year bond pays a 5-dollar coupon at the end of each year and 105 dollars at the end of year 2. Its annual benchmark yield is 4.00%, and its current credit spread is 1.00%. Assume deterministic cash flows and annual compounding, with the spread added to the benchmark yield. Calculate spread duration, estimate the price change for a 50-basis-point widening, and compare the estimate with a full repricing.
Method and initial price. The total yield used to discount cash flows is \(y+s=0.04+0.01=0.05\). The year-1 and year-2 present values are
Therefore, \(P=4.7619048+95.2380952=100.00\).
Duration. Apply the cash-flow formula, dividing each present value by \(1.05\) and multiplying by its payment time:
First-order estimate. A 50-basis-point widening is \(\Delta s=0.0050\). Thus
The estimated new price is \(100-0.9297052=99.0702948\), or $99.07 to the nearest cent.
Full repricing check. The new total yield is \(4.00\%+1.50\%=5.50\%\). Reprice both cash flows:
The repriced change is \(99.0768401-100=-0.9231599\), or a loss of $0.92 to the nearest cent. The duration estimate is close, but not identical, because it uses a local first-order approximation. The common mistake this method avoids is treating the benchmark yield as if it also changed: only the spread rose by 50 basis points.
Finite-Difference Spread Duration
A closed-form derivative is not always convenient. Spread duration can instead be estimated by repricing the instrument at nearby spreads while keeping the benchmark curve and other assumptions unchanged. With a central bump of size \(h\),
This expression approximates the negative derivative of price divided by current price. The lower-spread price appears first because, for a typical bond, it is the higher price. The two-sided calculation generally reduces the directional bias of using only an upward or downward bump, but it remains a numerical approximation. The bump \(h\) must be recorded in decimal rate units.
Worked Example: Estimate Spread Duration From Repricing
A hypothetical bond has a current price of $101.00 at its current spread. A pricing model returns $103.20 when the spread is lowered by 25 basis points and $99.10 when the spread is raised by 25 basis points. The benchmark yield curve is unchanged in both scenarios, and the bond’s deterministic cash flows and pricing assumptions are otherwise identical. Estimate spread duration and the price impact of a further 60-basis-point widening from the current spread.
Method. Use central finite differences with \(h=25\) basis points \(=0.0025\). The price difference across the two repricings is \(103.20-99.10=4.10\), and the denominator is
Therefore,
For a further 60-basis-point widening, \(\Delta s=0.0060\). The estimated price change is
The approximate new price is $96.08. This calculation uses prices from the same valuation model with the benchmark curve held fixed. A common mistake is to divide by the 25-basis-point one-sided bump rather than the full 50-basis-point separation between the two repriced spreads; the central-difference denominator is \(2h\).
Portfolio Spread Sensitivity
For a portfolio of positions whose spread shocks are assumed to be the same, the portfolio’s spread duration is the market-value-weighted average of the individual durations. This follows by adding the individual first-order dollar price changes and dividing by total market value. Positions with different spreads or different shocks should be treated using their respective sensitivities and shocks rather than forcing one common portfolio shock.
Worked Example: Aggregate Spread Duration and Dollar Loss
A hypothetical insurer holds two fixed-income positions. Position A has market value $800,000 and spread duration 2.5 years. Position B has market value $1,200,000 and spread duration 5.0 years. The durations were calculated at the current spreads under the same annual-compounding convention. Estimate the portfolio’s spread duration and dollar price change for a parallel 30-basis-point widening, with the benchmark curve unchanged.
Step 1: Calculate total market value.
Step 2: Calculate the market-value-weighted duration.
Step 3: Apply the common spread shock. A 30-basis-point widening is \(0.0030\), so the estimated portfolio change is
As a check, the position-level estimates are \(-(800{,}000)(2.5)(0.0030)=-$6,000\) for A and \(-(1{,}200{,}000)(5.0)(0.0030)=-$18,000\) for B. Their sum is \(-$24,000\), matching the aggregate estimate. The common mistake avoided here is taking the simple average of 2.5 and 5.0; the larger position must receive the larger weight.
Interpretation and Common Traps
Spread duration is a sensitivity measure, not a forecast of default losses and not a probability of a credit event. It describes the price response to a specified change in the market spread under stated pricing assumptions. The spread itself may reflect multiple sources of compensation, and the same change in quoted spread need not have the same effect on every instrument.
- Mixing spread changes with benchmark-rate changes. In this calculation, benchmark rates are fixed. If they also move, their contribution is outside this spread-only estimate.
- Using the wrong units. Convert basis points to decimal rates before multiplying. For example, 60 basis points is \(0.0060\).
- Reversing the sign. With positive spread duration, widening has an approximate negative price effect; tightening has an approximate positive effect.
- Treating an estimate as a full repricing. The duration formula is a local, first-order approximation. For a larger shock, reprice the cash flows under the changed spread and compare with the estimate.
- Assuming all positions share one shock. A portfolio-weighted duration is directly useful for a common parallel spread change. Different issuer or sector spread movements require position-specific calculations.
Check Your Understanding
For each problem, assume deterministic cash flows where relevant and hold the benchmark yield curve fixed during spread changes.
- A three-year zero-coupon bond is priced using annual compounding. The benchmark yield is 3.00% and the credit spread is 2.00%. What is its spread duration under the stated flat-yield model? (Answer: \(3/1.05=2.8571\) years.)
- Two positions have market values $800,000 and $1,200,000 and spread durations 2.5 and 5.0 years, respectively. Estimate their combined dollar price change for a parallel 30-basis-point spread tightening. (Answer: approximately +$24,000.)
- A bond has price $100 and spread duration 4.2 years. Estimate its price change for a 40-basis-point widening and its approximate new price. (Answer: \(-$1.68\); approximate new price $98.32.)
- A pricing model gives prices $103.20 at a spread 25 basis points below current, $101.00 at current spread, and $99.10 at a spread 25 basis points above current. Estimate spread duration using a central difference. (Answer: \(4.10/[2(101.00)(0.0025)]=8.1188\) years.)
- A two-year bond pays $5 at year 1 and $105 at year 2. Its benchmark yield is 4.00% and its spread is 1.00%, with annual compounding. Calculate current price and spread duration using the cash-flow formula. (Answer: price $100.00; spread duration 1.8594 years.)