Translating Swaption Leverage into a Forward Swap Rate Model
Summary
The document asks how to carry a swaption sensitivity ratio from a short-rate model into a model driven by the forward swap rate. The answer defines the underlying swap’s annuity and uses the relationship between swap value and swap rate: the swap’s sensitivity to the rate equals its annuity. Applying the chain rule links rate sensitivity in the original representation to swap-rate sensitivity in the new one.
Substituting that relationship into the leverage ratio cancels the swap’s rate sensitivity, leaving swaption delta with respect to the forward swap rate divided by the annuity. The annuity is derived from the discount curve, while the swaption delta comes from the new model. This gives a practical translation when the swap value has the stated annuity-times-rate-difference form and the sensitivities are taken consistently; the note does not address complications from alternative swap conventions or model-specific risk mappings.
Key ideas
- Express the underlying swap value as annuity times the difference between the swap rate and fixed rate.
- The swap’s sensitivity to the swap rate is its annuity.
- Use the chain rule to relate sensitivities with respect to the short rate and the forward swap rate.
- The leverage ratio becomes forward swap rate delta of the swaption divided by the annuity.
- The annuity comes from discounting, while the swaption delta is supplied by the new model.
Tags
Full text
# Port a model dependent swaption sensitivity to a new model
# Port a model dependent swaption sensitivity to a new model
I have a short rate model in which I have (among others) the following metric (for leverages) for a swaption : $$L = \frac{\frac{\partial}{\partial r}V_0^{\textrm{Swaption}}}{\frac{\partial}{\partial r}V_0^{\textrm{Swap}}}$$ where $V_0^{\textrm{Swaption}}$ is the time $0$ swaption model price and $V_0^{\textrm{Swap}}$ the time $0$ model price of the underlying swap.
Now I am moving from this short rate model to a model where a dynamic is prescribed on the forward swap rate, imagine for instance a lognormal Black model on the forward swap rate, to make it simple. I would like to port the metric $L$ to the new model.
The quantity $\frac{\partial}{\partial r}V_0^{\textrm{Swap}}$ is, I guess, nothing but the annuity (a.k.a. level or pvpb) of the underlying swap, equal to $\sum_i \delta_i P_{0,T_i}^d$ and given by the spot discount curve (and interpolation).
But how to port the quantity $\frac{\partial}{\partial r}V_0^{\textrm{Swaption}}$ to the new model though ? Of course, I can write $$\frac{\partial}{\partial r}V_0^{\textrm{Swaption}} = \frac{\partial s_0}{\partial r} \frac{\partial}{\partial s}V_0^{\textrm{Swaption}}$$ where $\frac{\partial}{\partial s}V_0^{\textrm{Swaption}}$ is the delta in the new model (w.r.t. the swap rate $s$), but what to do with $\frac{\partial s_0}{\partial r}$ quantity ?
If the port is not possible, what would be an equivalent metric in the new model ?
## Answer by Olórin (score 1)
https://quant.stackexchange.com/a/44351
We note $A$ the annuity, so that $V^{swap} = A(s - K)$ so that $\frac{\partial V^{swap}}{\partial s} = A$. As the chain rule gives $$\frac{\partial V^{swap}}{\partial r} = \frac{\partial s}{\partial r} \frac{\partial V^{swap}}{\partial s}$$ we get that $$\frac{\partial s}{\partial r} = \frac{1}{A} \frac{\partial V^{swap}}{\partial r}$$ and as the chain rule gives also : $$\frac{\partial V^{swaption}}{\partial r} = \frac{\partial s}{\partial r} \frac{\partial V^{swaption}}{\partial s}$$ which combined with the previous equation gives $$\frac{\partial V^{swaption}}{\partial r} = \frac{1}{A} \frac{\partial V^{swap}}{\partial r} \frac{\partial V^{swaption}}{\partial s}$$ and diving both sides by $\frac{\partial V^{swap}}{\partial r}$ gives $$L = \frac{1}{A} \frac{\partial V^{swaption}}{\partial s}$$ where $A$ is given by the spot discount curve and $\frac{\partial V^{swaption}}{\partial s}$ by the new model.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.