Why a Forward Swap Rate Is a Martingale Under Its Annuity Measure
Summary
The document explains that a forward swap rate is a martingale under the risk-neutral measure associated with the swap annuity numeraire. It presents the rate as a ratio: the difference between zero-coupon bond values at the swap’s start and end dates divided by the annuity value. This numeraire choice supplies the stated martingale property.
The question asks whether a similar numeraire can make a bond’s yield a martingale. It offers no resolution, derivation, or evidence for that question, and the proposed concern about the nonlinear relationship between bond prices and yields remains a conjecture. The useful point is the connection between choosing a numeraire and obtaining a martingale for a tradable ratio; the document does not establish a corresponding result for bond yields.
Key ideas
- A forward swap rate is presented as a ratio of a bond-value difference to the swap annuity value.
- Under the risk-neutral measure associated with that annuity numeraire, the forward swap rate is a martingale.
- The document asks whether bond yield has an analogous numeraire but does not answer it.
- The concern that price curvature with respect to yield prevents such a result is posed as a conjecture.
Tags
Full text
# Bond yield: is it martingale with respect to risk-neutral probability measure of some numeraire? # Bond yield: is it martingale with respect to risk-neutral probability measure of some numeraire? Let $t$ mean current time, let $T_0, T_n$ mean two times such that $T_0\le T_n$, and let $y_t[T_0, T_n]$ mean the forward swap rate of a swap starting at $T_0$ and ending at $T_n$. (I am ignoring $T_0+2$ issues, and assume that the swap starts at $T_0$.) Then under the annuity numeraire $N_t = P_t[T_0, T_n]$, the forward swap rate $y_t[T_0, T_n]$ is a martingale under the risk-neutral measure associated with $N_t$. This follows from the fact that $y_t$ is a ratio of a portfolio of assets by $P_t[T_0, T_n]$. Indeed, $$ y_t[T_0, T_n] = (Z(t,T_0)-Z(t,T_n))/P_t[T_0,T_n], $$ where $Z(t, T_i)$ means the zero-coupon bond from $t$ to $T_i$. Is there a corresponding numeraire for the yield of a bond? I am guessing the answer is no under some mild assumptions because bonds are tradeable and their price has a non-zero second derivative with respect to yields, but cannot hack through the thicket of results at the moment. Thanks in advance, any help appreciated!
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.