Lognormality of rates and term structure models (Q4487014)

In this paper the authors propose a new model of the term structure of interest rates and to analyze some of its properties. \textit{D. Heath}, \textit{R. Jarrow} and \textit{A. Morton} [Econometrica, 60, 77-105 (1992; Zbl 0751.90009)] extended the results of \textit{J. K. Ho} and \textit{T. C. Lee} [J. Finance 41, 1011-1029 (1986)]. Their results lead in a natural way to the modelling of forward rates as two parameters random fields or equivalently, as infinite dimensional stochastic processes. \textit{M. Musiela} [Journées Internationales de Finance, IGR-AFFI, La Baule] reparametrized the HJM model and then the non-arbitrage argument yielded a stochastic partial differential equation for the reparametrized forward rates process which can be rewritten as a stochastic evolution equation in an infinite dimensional Hilbert space with immediate advantage that the resulting solution is a Markov process. This approach has been latter studied by \textit{A. Brace}, \textit{D. Gatarek} and \textit{M. Musiela} [Math. Finance 7, 127-155 (1997; Zbl 0884.90008)] where a model of LIBOR rates, consistent with market practice has been proposed and analysed but latter it was pointed out by \textit{M. Musiela} and \textit{M. Rutkowski} [Finance Stoch., 261-291 (1997; Zbl 0888.60037)] that this suffers a serious difficulty, the resulting instantaneous forward rates need not be positive at all times. The authors propose the following new model in this paper with lognormal type volatility structure which lead to positive forward rates. They follow \textit{Sandmann} and \textit{Sondermann} [The Review of Future Markets, 12, 392-423 (1993) and Tech. Rept. No. B-263, University of Bonn] and shift the volatility specification from the rates \(r(t,x)\) to the rates \(j(t,x)\) which satisfy \(e^{r(t,x)}= 1+j(t,x)\). The process \(\{j(t,x); t,x<O\}\) will satisfy \[ \begin{multlined} dj(t,x)= \left({\partial \over\partial x} j(t,x)+ j(t,x)\gamma (t,x)\int^x_0 {j(t,u) \over 1+j(t,u)}\gamma (t,u)du+ {1\over 2}{j^2(t,x) \over +J(t,x)}\bigr |\gamma(t,x) \bigr|^2 \right)dt+\\ +j(t,x) \gamma(t,x).\end{multlined} \] The family of rates \(\{j(t,x); t,x\leq O\}\) can be viewed as a stochastic process \(\{j(t)=j(t,.): t\leq O\}\) with values in a space of function of \(x\). It has been proved that it has unique positive solution. This model enjoys many desirable properties: Rates are bounded below and above by a lognormal process, allows to hedge caps, swaptions and other currency derivatives including the Eurodollar futures contract and there exists a stationary distribution for the forward rates process. This model has already been applied to the pricing of American interest rate options.