A no‑arbitrage continuous‑time framework that models the entire instantaneous forward rate curve f(t,T) by specifying its volatility structure; under a chosen pricing measure the forward drift is then determined by the volatility via the HJM drift condition, so arbitrage‑free dynamics for bond prices follow from the volatility specification together with the initial forward curve.

Principle

In HJM the volatility specification for forward rates is the primary modelling choice; the no‑arbitrage (drift) restriction is a functional relation derived from that volatility so that any admissible volatility field together with the initial curve generates arbitrage‑free bond dynamics.

Demonstration

Illustrative scenario → A modeller specifies a parametric forward‑volatility σ_f(t,T) and uses the HJM drift condition to compute the drift μ_f(t,T); consequence → the modeller obtains model dynamics for f(t,T) consistent with the observed initial forward curve and can price bond derivatives by projecting under the chosen measure.

Misapplication

Treating forward‑rate volatility and drift as independent calibration knobs. Why plausible → separate estimation procedures for volatilities and drifts are common; semantic error → in HJM the drift is constrained by volatility to avoid arbitrage; correct interpretation → permissible drifts are determined by the chosen volatility structure.

Consequence

Flexibility: HJM can represent a wide range of term‑structure behaviours and exactly fit an initial curve; practical consequence → high dimensionality and calibration difficulty, and an inconsistent volatility choice can introduce arbitrage if the drift condition is not enforced correctly.

Reversal

When the volatility is restricted to a finite‑factor parametric form the framework reduces to a finite‑factor model (which may be easier to implement) and the drift restriction simplifies; under measure changes (e.g., forward measures versus risk‑neutral) the drift representation differs though the underlying no‑arbitrage relation holds.

Boundary

Clearly within → continuous forward‑rate modelling under no‑arbitrage where volatility is the modelling primitive. Boundary case → finite‑factor HJM (parametric factorization of σ_f) that yields tractable state‑space representations. Clearly outside → short‑rate only models that do not model the full forward curve or discrete tenor LIBOR market models unless derived as specializations.

Semantic Tension

Generality and exact fit to initial curve ↔ Practical implementability and calibration (infinite‑dimensional object vs finite‑factor numerical tractability).

Synthesis

HJM reframes term‑structure modelling: choosing a forward‑rate volatility field implicitly chooses the drift and therefore the entire arbitrage‑free dynamics; practical modelling reduces to selecting a volatility representation with a tractable trade‑off between fidelity and implementability.