I'm trying to see why the Libor and Swap Market Models are incompatible with each other.
I have found numerous sources such as the book of Brigo and Mercurio but my main source is a paper of Jamshidian from 1997
(www.math.ku.dk/~rolf/teaching/PhDcourse/Jlibor.pdf)
in which on page 321 (page 29 in the pdf) he states that since $S_i= \frac{\prod_{j=i}^{n-1}(1+\delta_j L_j)-1}{\sum_{j=1}^{n-1}\delta_j\prod_{k=j+1}^{n-1}(1+\delta_k L_k)}$
where $L_j$ is the Libor rate for the period $[T_{j_-1},T_j]$ and $S_i$ the forward swap rate for the time period $[T_i,T_n]$. In the paper it is seen that there exist measures and corresponding BMs such that S and L are lognormal under their individual measure with deterministic volatilities.
Jamishidian then states that this can not be because they can not both have determinsitic volatilities.
This argument seems much much simpler compared to the argumentation in the book of Brigo and Mercurio or in other Papers from Ruttkowski and others. Why does this work though?
Thank you in advance for the help or suggestions on where to look, Mattias