Skip to content
All library documents

Callable Bond Duration Under Exercise Scenarios

Article Quant Q&A · Author: Sane

Summary

The document explains two practical ways to estimate duration and related small-shock risk measures for callable bonds, where issuer exercise can change the bond’s expected cash flows. One approach assumes the issuer exercises at the yield-to-worst call date and measures risk as if that scenario will occur. The other assigns probabilities to possible call or maturity outcomes and averages the scenario-specific risk measures using those probabilities.

A three-year bond with call dates is used to illustrate the alternatives, including a probability-weighted duration across call and maturity scenarios. The discussion says the methods are intended for small market-data perturbations, such as duration or DV01, and should not be used for stress scenarios such as value-at-risk calculations. It does not derive a closed-form Macaulay duration formula; instead, it describes scenario assumptions and leaves the probability estimates dependent on volatility assumptions and modeling.

Key ideas

  • Callable-bond duration depends on which exercise outcome is assumed.
  • A yield-to-worst approach treats the issuer’s worst-yield exercise scenario as certain for small risk shocks.
  • A probability-weighted approach averages risk measures across call and maturity scenarios.
  • Scenario probabilities require assumptions about interest-rate and credit-spread volatility.
  • These methods are presented for small risk perturbations, not stress testing.

Tags

Full text
# Macaulay Duration of a Callable Bond


# Macaulay Duration of a Callable Bond












I could not find any formula of Macaulay duration for a callable bond in the literature. Can anybody show how to derive it or give a reference where it is already obtained.

EDIT My goal is to find a closed form formula for callable bond. This paper derives modified duration for callable bond (see Appendix for derivation). The derivation is not clear to me, as the author does not specify what kind of yield, $Y$, is used to take the partial derivative of prices (YTM, YTC, YTW). Also, it is not Macaulay duration, but modified duration.

## Answer by Dimitri Vulis (score 2, accepted)

https://quant.stackexchange.com/a/77838

There are two commonly used approaches, which are OK for for duration/dv01, key rate duration / dv01 by tenor bucket, and other risk measures in which the risk scenario perturbs the market data only by a small amount. Neither should be used for stress scenarios, such as in VaR calculations.

1 find the yield to worst. Assume that the bond issuer would exercise to seek the worst yield. (Note also that for putable bonds, which are much less common that callable, we'd assume that the bond holder would seek the best yield.) Assume that a small bump to the market data would not affect the moneyness of the option. Calculate the risk measures under these assumptions.

2 making some assumptions about the volatilities of the market data, for each exercise scenario, calculate the probability of the scenario being realized, and the risk measures assuming this scenario, then use the probability-wwighted sum of the risk measures.

Numerical example. A bond matures in 3 years, but has Bermudan calls in 1 and 1 years. Label the scenarios:

$s=1$ the issuer calls the bond in 1 year

$s=2$ the issuer calls the bond in 2 years

$s=3$ the issuer doesn't call the bond, but lets it mature in 3 years

Approach 1: making up some numbers, the yields under each scenario are $y_s$, the YTW is realized if the bond is called in 1 year. Assume that this will happen, ignore the possibility of the bond not being called, use the risk measures of the 1-year bond.

Approach 2: making some assumptions about the volatilities of the interest rates and the bond issuer's credit spread, and probably using a trer, calculate that the probabilities of being called in 1 or 2 years are $p_1$ and $p_2$, resulting in durations $d_1$ and $d_2$ respectively, and the probability of letting the bind mature is $p_3$, use $d=\sum_{s=1}^3p_sd_s$.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.