Heath-Jarrow-Morton

Overview

Heath-Jarrow-Morton (1992) is not a single model but the framework containing every arbitrage-free term structure model driven by Brownian motions: it takes the entire instantaneous forward curve as the state variable and shows that, under the risk-neutral measure, the drift of every forward rate is completely determined by the volatility structure. This implementation covers the Gaussian (deterministic-volatility) subclass with any number of factors: the user supplies factor volatility functions and the library computes the no-arbitrage drift by quadrature and prices by Monte Carlo. Constant volatility recovers Ho-Lee and exponentially damped volatility recovers Hull-White, which is how the implementation is validated in the test suite.

Dynamics

The instantaneous forward curve \(f(t, T)\) evolves as

\[df(t, T) = \alpha(t, T)\, dt + \sum_{i=1}^{n} \sigma_i(t, T)\, dW_i(t), \qquad f(0, T) = f^M(0, T)\]

with independent Brownian drivers. Absence of arbitrage forces the HJM drift condition:

\[\alpha(t, T) = \sum_{i=1}^{n} \sigma_i(t, T) \int_t^T \sigma_i(t, u)\, du\]

The short rate is the left end of the curve, \(r(t) = f(t, t)\), and bond prices are

\[P(t, T) = \exp\!\left(-\int_t^T f(t, u)\, du\right)\]

Because the initial condition is the market forward curve, the model fits the observed curve exactly by construction.

Special cases

  • \(\sigma(t, T) = \sigma\) (constant): Ho-Lee dynamics.

  • \(\sigma(t, T) = \sigma\, e^{-\lambda (T - t)}\): Hull-White dynamics with mean reversion \(\lambda\).

Both are available as the HJM.constant_vol and HJM.exponential_vol constructors, and general (multi-factor) structures are given as a list of callables vol(t, T_array).

Discretization

Simulation uses a uniform grid \(0 = t_0 < \dots < t_N = T_{\max}\) that doubles as the maturity grid. Each Euler step advances every live forward:

\[f(t_{k+1}, T_j) = f(t_k, T_j) + \alpha(t_k, T_j)\, \Delta t + \sum_i \sigma_i(t_k, T_j)\, \sqrt{\Delta t}\; Z_i, \qquad Z_i \sim \mathcal N(0, 1)\]

with the drift integral evaluated by cumulative trapezoidal quadrature along the maturity axis each step. The realized short-rate path is the grid diagonal \(r(t_k) = f(t_k, t_k)\); bond prices at expiry integrate the simulated curve by the trapezoidal rule. The martingale identity \(\mathbb E\left[e^{-\int_0^T r\, dt}\right] = P^M(0, T)\) is enforced in the tests as a direct check of the drift condition.

Parameters

Parameter

Meaning

Constraint

curve

market discount curve (initial forwards)

—

vols

factor volatility spec: float, callable, or list of callables

positive

T_max

longest maturity on the grid and simulation horizon

\(T_{\max} > 0\)

n_steps

number of time steps

\(\geq 2\)

Usage

import numpy as np
from interest_rate_models import HJM

# One factor, Hull-White-like
hjm = HJM.exponential_vol(curve, sigma=0.01, lam=0.1, T_max=5.0, n_steps=100)

# Two custom factors
def level(t, T):
    return 0.008 * np.exp(-0.5 * (T - t))

def slope(t, T):
    return 0.004 * np.ones_like(T)

hjm2 = HJM(curve, [level, slope], T_max=5.0, n_steps=100)

hjm.mc_bond_price(3.0, n_paths=50_000, seed=1)     # == curve.discount(3.0) up to MC error
hjm.mc_bond_option(1.0, 3.0, strike=0.93, kind="call")
times, short_rates = hjm.simulate(n_paths=10_000, seed=1)

References

  • Heath, D., Jarrow, R., Morton, A. (1992). “Bond pricing and the term structure of interest rates: A new methodology for contingent claims valuation”, Econometrica, 60(1), 77-105.

  • Brigo, D., Mercurio, F. (2006). Interest Rate Models — Theory and Practice, 2nd ed., Springer, ch. 5.