Heath-Jarrow-Morton (1992) Forward Rate Lab

1. Instantaneous Forward Rates and the Initial Term Structure

Prior to Heath, Jarrow, and Morton (1992), interest rate modeling was dominated by short-rate models (e.g. Vasicek 1977, Cox-Ingersoll-Ross 1985, Ho-Lee 1986). A primary limitation was the difficulty of fitting the entire initial yield curve without calibrating artificial time-dependent parameters. HJM revolutionized mathematical finance by shifting the fundamental modeling variable to the instantaneous forward rate $f(t, T)$, defined as the rate contracted at time $t$ for riskless borrowing or lending over the infinitesimal interval $[T, T+dT]$:

Because the model takes the market initial forward curve $f(0, T)$ as an exogenous initial condition, it fits the initial bond price term structure $P(0, T)$ by construction.

2. The HJM No-Arbitrage Drift Condition

Let the forward rate $f(t, T)$ evolve according to a multi-factor Itô stochastic differential equation under the empirical measure $\mathbb{P}$ or risk-neutral measure $\mathbb{Q}$:

Under the risk-neutral measure $\mathbb{Q}$ (with the money market bank account $B(t) = \exp(\int_0^t r(s) ds)$ as numeraire), the discounted price of any zero-coupon bond $\tilde{P}(t, T) = P(t, T) / B(t)$ must be a local martingale. Applying Itô's Lemma to $P(t, T) = \exp(-\int_t^T f(t, u) du)$ and setting the drift of $\tilde{P}(t, T)$ to zero yields the celebrated HJM No-Arbitrage Drift Condition:

Key Insight: In the risk-neutral world, there are zero free drift parameters. The drift of forward rates across all maturities is completely and uniquely dictated by the volatility structure $\boldsymbol{\sigma}(t, T)$ alone.

3. Separable Volatility and Markovian State Realizations

In a generic HJM specification, the spot short rate $r(t) = f(t, t)$ is non-Markovian—its evolution depends on the entire past history of Brownian increments $\int_0^t \sigma(s, t) dW(s)$. However, Carverhill (1994) and Ritchken & Sankarasubramanian (1995) established that the system admits a finite-dimensional Markovian state space if and only if the volatility structure is exponentially separable:

This specification corresponds exactly to the extended Vasicek / Hull-White (1990) one-factor short-rate model, proving that Hull-White is a special Markovian subset of the universal HJM framework.

4. Contingent Claims: Bond Options, Caps, and Swaptions

The HJM Monte Carlo engine prices contingent claims by simulating forward curve paths, accumulating the stochastic money market numeraire $B(t_E)$, and evaluating discounted expectation:

Under deterministic Gaussian volatility, this matches the analytical Black-76 bond option formula with integrated volatility $\sigma_P = \frac{\sigma}{\lambda}(1 - e^{-\lambda(T_B - t_E)})\sqrt{\frac{1 - e^{-2\lambda t_E}}{2\lambda}}$, providing an exact numerical benchmark for Monte Carlo validation.