Skip to content

The Heath-Jarrow-Morton Framework

The Heath-Jarrow-Morton (HJM) framework, introduced in 1992, takes a fundamentally different approach to modeling interest rate dynamics. Rather than specifying the evolution of the short rate (as in Vasicek or Hull-White), the HJM framework directly models the evolution of forward rates—the entire forward curve as a stochastic process. This framework provides a principled way to ensure no-arbitrage is built into the model by construction and has become the theoretical foundation for modern interest rate modeling.

The Fundamental Insight

Before HJM, interest rate models typically specified short-rate dynamics directly, and then bond prices were derived from these dynamics. The HJM framework inverts this perspective: it starts with the observed forward rate curve and specifies how this entire curve evolves through time. The key innovation is that the HJM framework imposes a drift restriction—known as the no-arbitrage condition—that links the deterministic and stochastic components of forward rate evolution. This restriction automatically eliminates arbitrage opportunities without requiring a specific parametric model form.

The philosophical shift is profound. Instead of asking "what should the short rate do?", HJM asks "given the market's current expectations (encoded in the initial forward curve), what paths can the curve take without violating no-arbitrage?" This perspective is more aligned with how traders think about markets: they observe market prices and infer constraints on what paths are possible.

Forward Rates and Bond Prices

Before diving into the HJM framework, let's establish the relationship between forward rates and bond prices. The instantaneous forward rate \(f(t, T)\) represents the rate at which the market agrees to lend or borrow at time \(t\) for an infinitesimal period starting at time \(T\). The connection to zero-coupon bond prices is:

\[ \begin{equation} P(t, T) = \exp\left(-\int_t^T f(t, u)du\right), \label{eq:bond_from_forward} \end{equation} \]

Taking the partial derivative with respect to maturity \(T\):

\[ \begin{equation} f(t, T) = -\frac{\partial \ln P(t, T)}{\partial T}. \label{eq:forward_from_bond} \end{equation} \]

The short rate \(r(t)\) is the limiting case: \(r(t) = f(t, t)\), the forward rate for an infinitesimal borrowing period starting immediately.

The HJM Framework: General Specification

The HJM framework assumes that the forward rate curve evolves according to:

\[ \begin{equation} df(t, T) = \alpha(t, T)dt + \sum_{i=1}^m \sigma_i(t, T)dW_i(t), \label{eq:hjm_forward_sde} \end{equation} \]

where:

  • \(f(t, T)\) is the forward rate at time \(t\) for maturity \(T\)
  • \(\alpha(t, T)\) is the drift function (deterministic)
  • \(\sigma_i(t, T)\) are the volatility functions for \(m\) risk factors
  • \(W_i(t)\) are independent standard Brownian motions

The critical feature of HJM is what happens next: the no-arbitrage condition imposes a constraint on the drift \(\alpha(t, T)\). It turns out that \(\alpha(t, T)\) cannot be arbitrary—it must be related to the volatilities in a very specific way.

The No-Arbitrage Drift Restriction

The no-arbitrage condition in HJM states that the drift of the forward rate is completely determined by its volatility structure:

\[ \begin{equation} \alpha(t, T) = \sum_{i=1}^m \sigma_i(t, T)\left(\sum_{j=1}^m \rho_{ij}\int_t^T \sigma_j(t, u)du\right), \label{eq:hjm_drift_restriction} \end{equation} \]

where \(\rho_{ij}\) is the correlation between the \(i\)-th and \(j-\)th Brownian motions. When the Brownian motions are independent (\(\rho_{ij} = 0\) for \(i \neq j\)), this simplifies to:

\[ \begin{equation} \alpha(t, T) = \sum_{i=1}^m \sigma_i(t, T)\int_t^T \sigma_i(t, u)du. \label{eq:hjm_drift_restriction_independent} \end{equation} \]

This is a profound result. It says that you cannot independently choose both drift and volatility; once you specify the volatility functions \(\sigma_i(t, T)\), the drift is automatically determined. This is the no-arbitrage drift restriction, and it ensures that the entire model is consistent with the absence of arbitrage opportunities.

Why Does This Work?

The intuition behind the drift restriction comes from the theory of bond pricing under the risk-neutral measure. Under the risk-neutral measure, all bond prices must grow at the short-rate (the risk-free rate). This constrains how forward rates can evolve. The HJM drift restriction emerges as a consequence of this requirement: the drift in forward rates along the curve must be such that bond prices, when expressed in terms of forward rates, grow at the short rate exactly.

Consider bonds of different maturities. Each has a risk-neutral drift equal to the short rate. Forward rates for different maturities are related through the bond pricing equation. If the forward rates evolve too quickly (or too slowly) at certain maturities, bond prices won't grow at the short rate as they should. The drift restriction ensures consistency across all maturities simultaneously.

One-Factor HJM Models

When we restrict to a single stochastic factor (\(m=1\)), the HJM framework becomes much more tractable. Many practical one-factor HJM models have been developed, and some of them are actually equivalent to well-known short-rate models we already know.

Gaussian HJM Models

If we specify a deterministic volatility function \(\sigma(t, T)\) and assume no jumps, we get a Gaussian HJM model. The entire forward curve evolves according to:

\[ \begin{equation} df(t, T) = \sigma(t, T)\int_t^T \sigma(t, u)du\, dt + \sigma(t, T)dW(t). \label{eq:gaussian_hjm} \end{equation} \]

Different choices of \(\sigma(t, T)\) lead to different models. For instance, if we choose \(\sigma(t, T) = \sigma e^{-a(T-t)}\) (exponential decay with maturity), the resulting short-rate dynamics become exactly the Hull-White model—establishing a direct connection between the HJM framework and the models we've seen before.

Connection to Hull-White Model

The Hull-White model emerges naturally as a one-factor Gaussian HJM model. To see this, consider the HJM framework with exponential volatility:

\[ \begin{equation} \sigma(t, T) = \sigma e^{-a(T-t)}, \label{eq:hjm_exponential_volatility} \end{equation} \]

Substituting into the no-arbitrage drift restriction:

\[ \begin{equation} \alpha(t, T) = \sigma e^{-a(T-t)} \int_t^T \sigma e^{-a(u-t)}du = \sigma^2 e^{-a(T-t)}\int_0^{T-t} e^{-as}ds = \frac{\sigma^2}{a}\left(1 - e^{-a(T-t)}\right)e^{-a(T-t)}. \label{eq:hjm_alpha_exponential} \end{equation} \]

The forward rate dynamics become:

\[ \begin{equation} df(t, T) = \frac{\sigma^2}{a}\left(1 - e^{-a(T-t)}\right)e^{-a(T-t)}dt + \sigma e^{-a(T-t)}dW(t). \label{eq:hjm_forward_dynamics} \end{equation} \]

Now the key step: integrate this forward rate evolution to get the short-rate dynamics. The short rate is \(r(t) = f(t, t)\), so:

\[ \begin{equation} dr(t) = df(t, t) = \frac{\sigma^2}{a}\left(1 - 1\right)e^{0}dt + \sigma e^{0}dW(t) = \sigma dW(t). \label{eq:hjm_to_shortrate_first} \end{equation} \]

This doesn't immediately give us Hull-White. We need to apply Itô's lemma more carefully to the path of the forward curve. When we do this calculation properly (integrating the forward rate differential across the curve), we obtain:

\[ \begin{equation} dr(t) = \left[\theta(t) - ar(t)\right]dt + \sigma dW(t), \label{eq:hull_white_from_hjm} \end{equation} \]

where \(\theta(t)\) is the drift function that must be calibrated to the market yield curve. This is exactly the Hull-White model! The function \(\theta(t)\) in the Hull-White model encodes the initial forward curve from the HJM perspective—it ensures that the short-rate model, when evolved, will generate bond prices consistent with the market's current expectations.

The Deeper Relationship

The relationship between HJM and Hull-White reveals something fundamental about interest rate modeling:

Hull-White can be viewed as a short-rate model that is derived from an underlying HJM model. When we specify exponential volatility in the HJM framework, the model is completely determined (the drift is given by the no-arbitrage condition), and when we look at the implied short-rate dynamics, they follow Hull-White with \(\theta(t)\) determined by the initial forward curve.

Conversely, every Hull-White model is consistent with some HJM model. The parameters \(a\) and \(\sigma\) determine the shape of volatility across maturities. The function \(\theta(t)\) is then derived from the market curve, ensuring the model is arbitrage-free.

This duality reflects a deep principle in financial mathematics: there are multiple ways to parameterize the same market. The HJM framework is more general (it can incorporate many volatility structures), while the Hull-White model is more practical (it has closed-form formulas and is easier to implement). The Hull-White model represents the most natural choice within the HJM framework when we want analytical tractability and parsimony of parameters.

Multi-Factor HJM Models

The HJM framework shines when modeling multiple factors. For instance, a two-factor HJM model might specify:

\[ \begin{equation} df(t, T) = \alpha(t, T)dt + \sigma_1(t, T)dW_1(t) + \sigma_2(t, T)dW_2(t), \label{eq:hjm_two_factor} \end{equation} \]

The drift restriction now becomes:

\[ \begin{equation} \alpha(t, T) = \sigma_1(t, T)\int_t^T \sigma_1(t, u)du + \sigma_2(t, T)\int_t^T \sigma_2(t, u)du, \label{eq:hjm_drift_two_factor} \end{equation} \]

assuming the two factors are independent. Two-factor models can capture both parallel shifts (level changes) and twists (slope changes) in the yield curve—something single-factor models cannot do. The cost is increased computational complexity; while one-factor HJM can sometimes admit closed-form bond prices, two-factor and higher models typically require numerical methods like Monte Carlo simulation.

Advantages and Limitations of HJM

Strengths of the HJM Framework:

The HJM framework provides a theoretically pure foundation for interest rate modeling. By specifying forwards instead of short rates, the entire term structure evolution is intrinsically no-arbitrage by construction. The framework is flexible: you can incorporate multiple factors, time-varying parameters, and even jumps. It also provides a principled way to think about the term structure—the market's expectations about all future rates are encoded in the initial forward curve and its stochastic evolution.

Limitations of the HJM Framework:

Despite its theoretical elegance, the HJM framework has practical challenges. First, implementing a full HJM model typically requires Monte Carlo simulation because most multi-factor specifications don't have closed-form solutions for bond prices. This is computationally expensive compared to short-rate models like Hull-White. Second, many HJM implementations can generate unrealistic forward rate paths—for instance, very long-dated forward rates might become negative or exhibit explosive behavior if volatility is not carefully chosen. Finally, calibrating the volatility functions \(\sigma_i(t, T)\) to market data is more difficult than calibrating a few parameters in a short-rate model; you must estimate an entire function, which requires rich data and careful statistical techniques.

HJM in Practice

In practice, the HJM framework is often used in its most tractable forms. The one-factor Gaussian HJM model (which is equivalent to Hull-White) remains the industry standard for many applications because it balances flexibility and tractability. Some practitioners use specific HJM models, such as those with exponential volatility or piecewise-constant parameters, to achieve particular properties while maintaining some analytical power.

For exotic derivatives and risk management, practitioners often work with the HJM framework conceptually—thinking about how the entire curve evolves—but implement it via a parsimonious short-rate model like Hull-White. This pragmatic approach leverages the theoretical insights of HJM while keeping computational costs manageable.

Summary: HJM and Hull-White

The relationship between HJM and Hull-White is not that Hull-White is derived from HJM, but rather that both are different perspectives on the same underlying reality. HJM provides a general framework for modeling the evolution of the entire forward curve while respecting no-arbitrage. When we impose specific forms of volatility and Gaussian dynamics within the HJM framework, we recover the Hull-White model as a natural special case. Conversely, every Hull-White model can be understood as implementing a specific HJM model with exponential volatility.

In this sense, Hull-White represents the optimal balance within the HJM framework for practitioners seeking both market consistency and analytical tractability. The framework explains why Hull-White works so well: it embodies the no-arbitrage drift restriction in a simple, implementable form, allowing practitioners to focus on calibrating the few parameters (\(a\) and \(\sigma\)) that truly matter while ensuring the entire model remains consistent with market prices and is free of arbitrage.

References

1 - The original paper introducing time-dependent parameters for yield curve fitting.

2 - Chapter 3.5: The Heath-Jarrow-Morton framework and its connection to short-rate models.

3 - Chapter 6: Detailed treatment of HJM models and market models.

Building confidence through rigorous validation

  1. John Hull and Alan White. Pricing interest-rate-derivative securities. The Review of Financial Studies, 3(4):573–592, 1990. doi:10.1093/rfs/3.4.573. ↩

  2. Damiano Brigo and Fabio Mercurio. Interest Rate Models – Theory and Practice: With Smile, Inflation and Credit. Springer, 2nd edition, 2006. URL: https://link.springer.com/book/10.1007/978-3-540-34604-3. ↩

  3. Leif Andersen and Vladimir Piterbarg. Interest Rate Modeling. Volume 1-3. Atlantic Financial Press, 2010. ↩