BGM LIBOR Market Model Lab | Interest Rate Derivatives

Q: What is the Brace-Gatarek-Musiela (BGM / LMM) LIBOR Market Model?

The Brace-Gatarek-Musiela (1997) LIBOR Market Model (BGM / LMM) is an institutional multi-factor interest rate modeling framework that simulates observable discrete forward rates directly rather than unobservable instantaneous short rates. Because forward rates are lognormal martingales under their associated forward measures, market-standard Black-76 caplet prices calibrate exactly by definition.

Q: Why does the BGM model price standard caplets identically to the Black-76 formula?

Under the specific forward risk-neutral measure Q^{k+1} where the zero-coupon bond P(t, T_{k+1}) serves as the numeraire, the forward rate L_k(t) has zero drift and evolves as a driftless geometric Brownian motion. Consequently, the discounted expectation of the caplet payoff yields the analytical Black-76 formula in closed form.

Q: How does the Hunter-Jäckel-Joshi predictor-corrector drift stepping work?

Under a single common measure such as the Spot LIBOR measure, forward rates have state-dependent drifts that depend on other forward rates. The Hunter, Jäckel, and Joshi (2001) predictor-corrector scheme computes an initial Euler drift forecast (predictor), evaluates terminal rates, recalculates drift at the new state, and uses the arithmetic average drift (corrector) to remove log-Euler discretization bias.

Q: What is Rebonato's formula for analytical European swaption volatility?

Rebonato's formula provides a closed-form approximation for European swaption implied volatility by expressing the forward swap rate as a linear combination of underlying forward LIBOR rates. By freezing weights at t=0, the total swaption variance is computed as a double summation over forward rate variances, volatilities, and cross-tenor correlations.

Q: Why is a multi-factor market model required to price path-dependent ratchet caps?

A ratchet cap features coupon strikes that reset to the previously realized forward rate plus a spread, introducing joint path dependency across consecutive tenors. One-factor short rate models enforce perfect inter-tenor correlation (rho=1), causing severe mispricing of curve slope shocks. BGM's multi-factor correlation structure captures yield curve twists, slope shifts, and curvature dynamics accurately.

Market Regimes:

Canonical 5-year semi-annual forward curve flat at 4.0%, Rebonato humped volatility peaking at ~1.5 years, 2-factor correlation.

Mathematical Foundations: Brace-Gatarek-Musiela (BGM / LMM)

1. Discrete Forward Rate SDE Dynamics

Consider a discrete tenor structure $0 = T_0 < T_1 < \dots < T_N$ with spans $\tau_k = T_{k+1} - T_k$. The simply compounded forward rate spanning $[T_k, T_{k+1}]$ is defined by:

$$L_k(t) = L(t; T_k, T_{k+1}) = \frac{P(t, T_k) - P(t, T_{k+1})}{\tau_k P(t, T_{k+1})}$$

Under its own forward measure $\mathbb{Q}^{k+1}$ (with numeraire bond $P(t, T_{k+1})$), $L_k(t)$ is a martingale with zero drift:

$$\frac{dL_k(t)}{L_k(t)} = \sigma_k(t) \, dW_t^{k+1}$$

This guarantees that individual caplet payoffs $\tau_k (L_k(T_k) - K)^+$ recover the standard market Black-76 formula in closed form.

2. Drift Adjustments under Spot LIBOR Measure

To simulate all forward rates simultaneously along a single path, a common numeraire is required. Under the discrete Spot LIBOR measure $\mathbb{Q}^*$ with rolling money market numeraire $B^*(t)$, the forward rate dynamics acquire state-dependent drift adjustments:

$$\frac{dL_k(t)}{L_k(t)} = \left( \sum_{j=m(t)}^k \frac{\tau_j L_j(t) \sigma_k(t) \sigma_j(t) \rho_{k,j}}{1 + \tau_j L_j(t)} \right) dt + \sigma_k(t) \sum_{p=1}^d v_{k,p} dW_t^{(p)}$$

3. Hunter-Jäckel-Joshi Predictor-Corrector Scheme

Because drift $\mu_k(L(t))$ depends nonlinearly on the state vector $L(t)$, standard Euler log-stepping introduces significant discretization bias. The predictor-corrector algorithm removes this bias:

Predictor: Compute drift $\mu_k(t_s)$ at current state $L(t_s)$ and project candidate rates $\tilde{L}(t_{s+1})$.
Corrector: Evaluate drift $\mu_k(\tilde{L}(t_{s+1}))$ at the candidate state.
Average Step: Advance the state using the average drift $\frac{1}{2}[\mu_k(t_s) + \mu_k(\tilde{L}(t_{s+1}))]$.

4. Rebonato Swaption Implied Volatility Formula

Forward swap rates $S_{n,m}(t) = \sum_{j=n}^{m-1} w_j(t) L_j(t)$ can be approximated analytically by freezing the annuity weights $w_j(0) = \frac{\tau_j P(0, T_{j+1})}{\sum_{l=n}^{m-1} \tau_l P(0, T_{l+1})}$:

$$\sigma_{\text{swap}}^2 \approx \sum_{j=n}^{m-1} \sum_{k=n}^{m-1} \frac{w_j(0) w_k(0) L_j(0) L_k(0) \rho_{j,k}}{S_{n,m}(0)^2 T_n} \int_0^{T_n} \sigma_j(t) \sigma_k(t) \, dt$$

Interactive Mastery Quiz: BGM / LIBOR Market Model

Test your knowledge of forward rate measures, predictor-corrector drift stepping, and swaption calibration.

1. Why is the forward rate $L_k(t)$ a martingale without drift under the measure $\mathbb{Q}^{k+1}$?

Because interest rates can never become negative under physical measure.

Because the money market bank account $B(t)$ is assumed to be deterministic.

Under numeraire bond $P(t, T_{k+1})$, the asset-to-numeraire ratio $[P(t, T_k) - P(t, T_{k+1})]/P(t, T_{k+1})$ is a martingale by Girsanov's theorem.

Because the drift is absorbed by the short rate instantaneous variance.

2. What major practical advantage did the BGM / LMM model provide over traditional short-rate models like Vasicek or Hull-White?

It eliminated the need for random number generators in Monte Carlo pricing.

It directly models discrete observable forward rates, calibrating exactly to quoted market Black-76 caplets.

It models the stock index return directly rather than interest rates.

It restricts all yield curves to have constant zero slope.

3. What is the fundamental purpose of the Hunter-Jäckel-Joshi predictor-corrector discretization algorithm?

To force all simulated interest rates to stay strictly below 10%.

To invert the correlation matrix using Fast Fourier Transform.

To compute American option early exercise boundaries without regression.

To average drift across time intervals, removing log-Euler numerical discretization bias across large step sizes.

4. How does Rebonato's formula approximate European swaption implied volatility?

By freezing forward rate annuity weights at t=0 and expressing swap rate variance as a double sum over forward covariances.

By solving a cubic algebraic equation for the critical stopping boundary.

By evaluating cosine series expansion terms on a truncated cumulant interval.

By assuming forward rates have zero correlation across all maturities.

Frequently Asked Questions

The Brace-Gatarek-Musiela (1997) LIBOR Market Model (BGM / LMM) is an institutional multi-factor interest rate modeling framework that simulates observable discrete forward rates directly rather than unobservable instantaneous short rates. Because forward rates are lognormal martingales under their associated forward measures, market-standard Black-76 caplet prices calibrate exactly by definition.

Under the specific forward risk-neutral measure Q^{k+1} where the zero-coupon bond P(t, T_{k+1}) serves as the numeraire, the forward rate L_k(t) has zero drift and evolves as a driftless geometric Brownian motion. Consequently, the discounted expectation of the caplet payoff yields the analytical Black-76 formula in closed form.

Under a single common measure such as the Spot LIBOR measure, forward rates have state-dependent drifts that depend on other forward rates. The Hunter, Jäckel, and Joshi (2001) predictor-corrector scheme computes an initial Euler drift forecast (predictor), evaluates terminal rates, recalculates drift at the new state, and uses the arithmetic average drift (corrector) to remove log-Euler discretization bias.

Rebonato's formula provides a closed-form approximation for European swaption implied volatility by expressing the forward swap rate as a linear combination of underlying forward LIBOR rates. By freezing weights at t=0, the total swaption variance is computed as a double summation over forward rate variances, volatilities, and cross-tenor correlations.

A ratchet cap features coupon strikes that reset to the previously realized forward rate plus a spread, introducing joint path dependency across consecutive tenors. One-factor short rate models enforce perfect inter-tenor correlation (rho=1), causing severe mispricing of curve slope shocks. BGM's multi-factor correlation structure captures yield curve twists, slope shifts, and curvature dynamics accurately.

Brace-Gatarek-Musiela (BGM / LMM) LIBOR Market Model Lab

Yield Curve & Market Dynamics

Rebonato Humped Volatility: $[a + b\tau]e^{-c\tau} + d$

Derivatives Specifications

Model Calibration Diagnostics

High-Dimensional Term Structure Visualizer

Swaption Surface: Strike ($K$) vs. Volatility Scale

Exotic Ratchet Cap: Correlation ($\beta$) vs. Factors ($d$)

Mathematical Foundations: Brace-Gatarek-Musiela (BGM / LMM)

1. Discrete Forward Rate SDE Dynamics

2. Drift Adjustments under Spot LIBOR Measure

3. Hunter-Jäckel-Joshi Predictor-Corrector Scheme

4. Rebonato Swaption Implied Volatility Formula

Interactive Mastery Quiz: BGM / LIBOR Market Model

Frequently Asked Questions

Related Fixed Income & Interest Rate Derivatives Labs

Hull-White One-Factor Lab

European Swaption Pricing Lab

Bermudan Swaption Lab

Interest Rate Cap & Floor Lab