Bond Price Ratios Are Martingales Under the Forward Measure
Summary
The document explains why the ratio of two zero-coupon bond prices, with maturities S and T where S is earlier than T, is a martingale under the T-forward measure. This property follows from the numeraire choice that defines the measure, so it does not depend on assuming a particular short-rate model such as CIR.
For a model-based verification exercise, the answer suggests deriving bond prices under the risk-neutral measure, finding the change of measure to the T-forward measure, and then establishing the ratio’s martingale property. A second derivation uses conditional expectations and the risk-neutral pricing relation for discounted bond prices to recover the same ratio at an earlier time. The document offers a theoretical result and a proof strategy; it does not work through CIR-specific bond-price calculations or discuss practical estimation.
Key ideas
- Under the T-forward measure, the price of an S-maturity bond divided by the T-maturity bond price is a martingale.
- The result follows from the choice of numeraire and is independent of the short-rate model.
- A model-specific proof can use bond prices, the Radon–Nikodym change of measure, and the resulting dynamics.
- Conditional expectation and risk-neutral pricing of discounted bonds provide an alternative derivation.
Tags
Full text
# Is $\frac{P(t,S)}{P(t,T)}$ martingale?
# Is $\frac{P(t,S)}{P(t,T)}$ martingale?
Assume $r_t$ follow the CIR process and $P(t,T)=E[exp(-\int_{t}^{T}r_s ds)|F_t]$.I am going to show $\frac{P(t,S)}{P(t,T)}$ ($S<T$) is an $F_t$-martingale under Forward Measure but So confused! Do I need to solve C.I.R process? Should I use the definition of martingale? please guide me! so thanks.
## Answer by Gordon (score 3)
https://quant.stackexchange.com/a/19174
By definition of the $T$-forward measure $P_T$, the process $\Big\{\frac{P(t,S)}{P(t,T)} \mid t\geq 0\Big\}$ is a martingale under the measure $P_T$, without assuming any specific models of the short rate $r_t$. That is, this martingale property is model independent.
However, as a good exercise, you can also do the following:
- Given the CIR interest rate model under the risk-neutral measure $P$, compute the bond prices.
- Find the Radon-Nykodim derivative of the $T$-forward measure with respect to the risk-neutral measure, that is, $\frac{dP_T}{dP}\big|_t$, for $0 \leq t \leq T$.
- Find the bond price formula or SDE under the $T$-forward measure.
- Show that the process $\Big\{\frac{P(t,S)}{P(t,T)} \mid t\geq 0\Big\}$ is a martingale under the $T$-forward measure.
## Answer by user16891 (score 3)
https://quant.stackexchange.com/a/19184
We assume $\mathbb{Q}$ is forward measure. $$\mathbb{E^Q}\left[\frac{P(t,S)}{P(t,T)}|\mathcal{F}_s\right]=\mathbb{E^P}\left[\frac{P(t,S)}{P(t,T)}\frac{e^{-\int_{s}^{T}r_u\,du}}{P(s,T)}\,|\,\mathcal{F}_s\right]$$ $$\hspace{5cm}=\frac{1}{P(s,T)}\mathbb{E^P}\left[\frac{P(t,S)}{P(t,T)}{e^{-\int_{s}^{T}r_u\,du}}\,|\,\mathcal{F}_s\right]$$ $$\hspace{6.9cm}=\frac{1}{P(s,T)}\mathbb{E^P}\left[\mathbb{E^P}\left[\frac{P(t,S)}{P(t,T)}{e^{-\int_{s}^{T}r_u\,du}}\,|\,\mathcal{F}_t\right]|\mathcal{F}_s\right]$$ we have $$\hspace{0.3cm}\mathbb{E^Q}\left[\frac{P(t,S)}{P(t,T)}|\mathcal{F}_s\right]=\frac{1}{P(s,T)}\mathbb{E^P}\left[{e^{-\int_{s}^{t}r_u\,du}}\frac{P(t,S)}{P(t,T)}\mathbb{E^P}\left[e^{-\int_{t}^{T}r_u\,du\,}\,\,|\,\mathcal{F}_t\right]|\mathcal{F}_s\right]$$ $$\hspace{1.5cm}=\frac{1}{P(s,T)}\mathbb{E^P}\left[{e^{-\int_{s}^{t}r_u\,du}}\frac{P(t,S)}{P(t,T)}P(t,T)|\mathcal{F}_s\right]$$
$$=\frac{1}{P(s,T)}\mathbb{E^P}\left[{e^{-\int_{s}^{t}r_u\,du}}P(t,S)|\mathcal{F}_s\right]$$ then $$\mathbb{E^Q}\left[\frac{P(t,S)}{P(t,T)}|\mathcal{F}_s\right]=\frac{1}{P(s,T)}{e^{\int_{0}^{s}r_u\,du}}\,\,\mathbb{E^P}\left[{e^{-\int_{0}^{t}r_u\,du}}P(t,S)|\mathcal{F}_s\right]$$ we know the discounted bond price process $\{e^{-\int_{0}^{t}r_u\,du}P(t,S)\}$ is a martingale under $\mathbb{P}$, thus we have $$\hspace{1cm}\mathbb{E^Q}\left[\frac{P(t,S)}{P(t,T)}|\mathcal{F}_s\right]=\frac{1}{P(s,T)}{e^{\int_{0}^{s}r_u\,du}}{e^{-\int_{0}^{s}r_u\,du}}P(s,S)$$ $$=\frac{P(s,S)}{P(s,T)}$$
$$$$ EDIT: Alternatively, \begin{align*} \mathbb{E^Q}\left[\frac{P(t,S)}{P(t,T)}\mid \mathcal{F}_s\right] &= \frac{1}{P(s,T)}\mathbb{E^P}\left[{e^{-\int_{s}^{t}r_u\,du}}P(t,S)\mid\mathcal{F}_s\right]\\ &=\frac{1}{P(s,T)}\mathbb{E^P}\left[{e^{-\int_{s}^{t}r_u\,du}}\mathbb{E^P}\Big(e^{-\int_t^Sr_u\,du} \mid \mathcal{F}_t \Big)\mid\mathcal{F}_s\right]\\ &=\frac{1}{P(s,T)}\mathbb{E^P}\left[\mathbb{E^P}\Big(e^{-\int_s^Sr_u\,du} \mid \mathcal{F}_t \Big)\mid\mathcal{F}_s\right]\\ &= \frac{1}{P(s,T)}\mathbb{E^P}\left[e^{-\int_s^Sr_u\,du}\mid\mathcal{F}_s\right]\\ &= \frac{P(s,S)}{P(s,T)}. \end{align*}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.