LIBOR Market Model¶
Overview¶
The LIBOR Market Model (Brace-Gatarek-Musiela 1997; also BGM, or the lognormal forward-LIBOR model) abandons the unobservable short rate and models what the market actually quotes: simply-compounded forward rates over a discrete tenor structure. Each forward is lognormal under its own forward measure, so caplets price by Black’s formula exactly — the model is the market’s cap quotation convention. The joint dynamics under a single (spot) measure carry state-dependent drifts and are simulated by a log-Euler scheme; European swaptions price approximately by Rebonato’s formula and exactly by Monte Carlo.
Setup and dynamics¶
Fix a tenor \(\tau\) and dates \(T_i = i\tau\), \(i = 0, \dots, N\). The state is the vector of forwards
each living until its fixing date \(T_i\). Under the \(T_{i+1}\)-forward measure, \(F_i\) is a driftless geometric Brownian motion with volatility \(\sigma_i\). Under the spot measure (the discretely rebalanced bank account \(B_d\)), the dynamics are
where \(\eta(t)\) is the index of the first forward not yet fixed and the driving Brownian motions are correlated as
The numeraire compounds period by period:
Caplet pricing¶
A caplet with rate \(K\) on forward \(i\) fixes at \(T_i\), pays \(\tau \max(F_i(T_i) - K,\, 0)\) at \(T_{i+1}\), and prices by Black’s formula:
A cap is the sum of its caplets (the first period, whose rate is already fixed, is excluded by convention).
Swap rates and swaptions¶
The par rate and annuity of a swap spanning \([T_\alpha, T_\beta]\):
Rebonato’s approximation freezes the weights \(w_i = \tau P(0, T_{i+1}) / A_{\alpha\beta}(0)\) in the exact identity \(S_{\alpha\beta} = \sum_{i=\alpha}^{\beta-1} w_i F_i\) and reads off an effective Black volatility for the swap rate:
The payer swaption then prices as Black on the annuity numeraire:
The Monte Carlo pricer reconstructs the curve at expiry from the simulated forwards, \(P(T_\alpha, T_j) = \prod_{l=\alpha}^{j-1} \left(1 + \tau F_l(T_\alpha)\right)^{-1}\), and discounts the annuity-weighted payoff with the bank account; the test suite checks the two agree.
Simulation¶
Log-Euler with the drift recomputed every substep:
Forwards freeze at their fixing dates. Correlated increments come from a PSD square root of \(\rho\) (eigendecomposition, so \(\beta = 0\) — perfectly correlated one-factor dynamics — is supported).
Parameters¶
Parameter |
Meaning |
Constraint |
|---|---|---|
|
market discount curve (initial forwards) |
— |
|
accrual period \(\tau\) in years |
\(\tau > 0\) |
|
number of forwards \(N\) |
\(\geq 2\) |
|
Black volatility per forward (scalar or length-\(N\) array) |
positive |
|
correlation decay \(\beta\) |
\(\beta \geq 0\) |
Usage¶
from interest_rate_models import LiborMarketModel
lmm = LiborMarketModel(curve, tenor=0.5, n_periods=10, vol=0.20, corr_beta=0.1)
lmm.caplet_price(4, strike=0.035)
lmm.cap_price(strike=0.035, maturity=3.0)
lmm.swap_rate(2, 6)
lmm.swaption_price(alpha=2, beta=6, strike=0.035) # Rebonato
lmm.mc_swaption_price(alpha=2, beta=6, strike=0.035, n_paths=100_000, seed=1)
dates, forwards, numeraire = lmm.simulate(n_paths=50_000, seed=1)
References¶
Brace, A., Gatarek, D., Musiela, M. (1997). “The market model of interest rate dynamics”, Mathematical Finance, 7(2), 127-155.
Rebonato, R. (2002). Modern Pricing of Interest-Rate Derivatives, Princeton University Press.
Brigo, D., Mercurio, F. (2006). Interest Rate Models — Theory and Practice, 2nd ed., Springer, chs. 6-8.