Proving Forward Rate Martingality with a Forward Measure
Summary
The document explains why a forward rate is a martingale under the measure associated with the bond maturing at the end of the accrual period. It defines the money-market and bond numeraires, then gives the Radon–Nikodym relationship between two forward measures and applies Bayes’ rule to change measures in a conditional expectation.
The proof reduces the forward rate to a ratio of zero-coupon bond prices, since the rate is an affine transformation of that ratio. Applying the measure-change formula makes the bond terms cancel inside the expectation, leaving the current bond-price ratio and establishing the martingale result. A second explanation gives the intuition: the forward times the relevant zero-coupon bond can be treated as a traded asset, whose price divided by its numeraire is a martingale under the associated measure. The argument relies on the stated pricing and measure framework; the document offers algebra and intuition but no numerical example.
Key ideas
- A forward measure is associated with a zero-coupon bond used as numeraire.
- Bayes’ rule and the Radon–Nikodym derivative connect expectations under different forward measures.
- The forward rate is an affine transformation of a ratio of bond prices.
- Under the terminal bond measure, the bond ratio is a martingale, which establishes the result for the forward rate.
- The forward rate multiplied by the relevant bond can be viewed as a traded asset for the numeraire argument.
Tags
Full text
# Forward Rates are martingal under Forwar Measure detailled proof
# Forward Rates are martingal under Forwar Measure detailled proof
So I read an other post about this :
How to prove martingality of forward rate under T-forward measure
But I can't see how to get from there to there :
$F \left(t,T_n \right)P \left(t,T_{n+1}\right) = \frac{1}{\tau} \left(P \left(t,T_{n}\right)-P \left(t,T_{n+1}\right)\right)$
$\frac{F \left(t,T_n \right)P \left(t,T_{n+1}\right)}{P \left(t,T_{n+1}\right)}=E^{T} \left[ \left. \frac{F \left(S,T_n \right)P \left(S,T_{n+1}\right)}{P \left(S,T_{n+1}\right)} \right| \mathcal{F}_t\right]$
If anyone could explain in depth this passage I would be really glad.
Thanks
## Answer by AXH (score 1, accepted)
https://quant.stackexchange.com/a/49896
The measure $\mathbb{Q}$ is associated to the money market account $t \mapsto \beta_t = \exp \int_{0}^{t} r_s d s $. The measure $\mathbb{Q}^T$ is associated to the zero coupon bond $t \mapsto P_{tT}$ where $ P_{tT}:=\mathbb{E}^{\mathbb{Q}}_t \left[ \frac{\beta_t}{\beta_T} \right]$. We know that the Radon-Nikodym derivative between $\mathbb{Q}^{T_1}$ and $\mathbb{Q}^{T_2}$ is given by $$ t \mapsto \frac{d \mathbb{Q}^{T_2}}{d\mathbb{Q}^{T_1}}(t)=\frac{P_{tT_2}}{P_{0T_2}} \cdot \frac{ P_{0T_1} }{ P_{tT_1} } $$ We know by the Bayes theorem for a payoff function $X$ observed at time $T$ that $$ \mathbb{E}^{\mathbb{Q}^{T_2}}_{t} \left[ X_T \right] = \frac{\mathbb{E}^{\mathbb{Q}^{T_1}}_{t} \left[ X_T \cdot \frac{d \mathbb{Q}^{T_2}}{d\mathbb{Q}^{T_1}}(T) \right]}{\mathbb{E}^{\mathbb{Q}^{T_1}}_{t} \left[ \frac{d \mathbb{Q}^{T_2}}{d\mathbb{Q}^{T_1}}(T) \right]} $$ Equivalently, $$ \mathbb{E}^{\mathbb{Q}^{T_2}}_{t} \left[ X_T \right] = \frac{\mathbb{E}^{\mathbb{Q}^{T_1}}_{t} \left[ X_T \cdot \frac{d \mathbb{Q}^{T_2}}{d\mathbb{Q}^{T_1}}(T) \right]}{\frac{d \mathbb{Q}^{T_2}}{d\mathbb{Q}^{T_1}}(t)} $$ Now that our tools are ready, let us attack. To prove that the process $$ F(t;T_1,T_2)=\frac{1}{\tau} \left[ \frac{ P_{tT_1} }{ P_{tT_2} } -1 \right] $$ is a $\mathbb{Q}^{T_2}$ martingale, it is equivalent to prove the same statement for the ratio of bonds $ X_T :=P_{TT_1} / P_{T T_2}$. Proceed as follows: $$ \begin{align} \mathbb{E}^{\mathbb{Q}^{T_2}}_t \left[ \frac{ P_{TT_1} }{ P_{TT_2} } \right] & = \frac{\mathbb{E}^{\mathbb{Q}^{T_1}}_t \left[ \frac{d \mathbb{Q}^{T_2}}{d\mathbb{Q}^{T_1}}(T) \cdot \frac{ P_{TT_1} }{ P_{TT_2} } \right]}{ \frac{d \mathbb{Q}^{T_2}}{d\mathbb{Q}^{T_1}}(t) } \\ & = \frac{\mathbb{E}^{\mathbb{Q}^{T_1}}_t \left[ \frac{ P_{TT_2} P_{0T_1} }{ P_{TT_1} P_{0T_2} } \cdot \frac{ P_{TT_1} }{ P_{TT_2} } \right]}{ \frac{ P_{tT_2} P_{0T_1} }{ P_{tT_1} P_{0T_2} } } \\ & = \frac{ P_{tT_1}}{P_{tT_2} } \end{align} $$
## Answer by Magic is in the chain (score 3)
https://quant.stackexchange.com/a/49893
The first equation is the result of the effort to show that the product of the forward and the relevant zero coupon, $F \left(t,T_n \right)P \left(t,T_{n+1}\right)$, can be treated as a traded asset.
And once you have established that it is the price of a traded asset, then you can write its price using the valuation formula, which in way holds for all trades asset. It essentially says that the price of an asset, say V, divided by a numeraire asset, say B, will be a martingale under some probability measure.
$\frac{V_t}{B_t}=E^Q\left[\left. \frac{V_S}{B_S} \right|\mathcal{F}_t\right]$
So in the second equation you are just plugging in the F times P, i.e. $V_t=F \left(t,T_n \right)P \left(t,T_{n+1}\right)$, for the traded asset V, and choosing $P \left(t,T_{n+1}\right)$ as the numeraire (in place of $B_t$).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.