Skip to content

Economic Modelling · Models of the term structure of interest rates

Multifactor and Forward Rate Models: Two-Factor, HJM and LIBOR

Updated 11 October 2026 · Fact-checked

Multifactor and forward rate models describe the whole yield curve with more than one source of randomness. Two-factor models add a second short rate or long rate driver. The HJM framework models instantaneous forward rates directly, with drift fixed by no-arbitrage. LIBOR market models model observable forward LIBOR rates.

Understand Multifactor and Forward Rate Models

A one-factor short rate model has a single source of randomness, usually the short rate r(t). Every bond yield then moves because of that one shock. So all points on the yield curve are driven by the same shock and move in a perfectly or highly correlated way. Real curves do not behave like this. Short and long yields often move differently, and the curve can steepen, flatten or twist.

The fix is to add factors. In a two-factor short rate model you might have a short rate and a long-term mean level that is itself random. Or you might have two correlated stochastic components that add up to r(t). The model can then produce level and slope movements, and more realistic correlation between rates of different maturities. The cost is more parameters and harder calibration.

The Heath-Jarrow-Morton (HJM) framework takes a different route. Instead of modelling the short rate and deriving the curve, it models the whole instantaneous forward rate curve f(t,T) directly. The starting curve is an input, so the model fits today's market curve by construction. The key result is that under the risk-neutral measure the drift of f(t,T) is not free. It is fixed by the volatilities. Choose the volatilities and no-arbitrage gives you the drift.

The LIBOR market model (also called the BGM model) models a set of discrete forward LIBOR rates, which are quoted in the market. Each forward rate is lognormal under its own forward measure. This matches the Black formula used to price caps, so it fits market practice well. Its weakness is that it is high-dimensional and usually needs simulation for complex products.

For the exam, focus on why more than one factor helps, what HJM takes as input, and why drift is determined by volatility. You are not usually asked for long derivations. You are asked to explain, compare and interpret.

Key rules to remember

Instantaneous forward rate
f(t,T) = −∂ ln P(t,T) / ∂T
P(t,T) is the price at time t of a zero-coupon bond paying 1 at T.
Bond price from forward rates
P(t,T) = exp(−∫ f(t,s) ds), integrated from s = t to s = T
Gives the full price curve from the forward curve.
Short rate from forward rate
r(t) = f(t,t)
The short rate is the forward rate with zero time to maturity.
HJM forward rate dynamics
df(t,T) = α(t,T) dt + σ(t,T) dW(t)
One-factor form. With several factors σ dW becomes a sum over the factors.
HJM no-arbitrage drift (risk-neutral measure, one factor)
α(t,T) = σ(t,T) × ∫ σ(t,s) ds, integrated from s = t to s = T
The drift is fully determined by the volatility. This is the key HJM result.
Simple forward LIBOR rate
L(t; T, T+δ) = [P(t,T) ÷ P(t,T+δ) − 1] ÷ δ
δ is the accrual period, for example 0.25 years.

How to solve Multifactor and Forward Rate Models questions

Use this approach for descriptive and numerical questions on multifactor and forward rate models.

  1. 1Identify what is being modelled: the short rate, the instantaneous forward curve (HJM), or discrete forward LIBOR rates (market model).
  2. 2State the number of factors and what each factor represents, such as level or slope.
  3. 3Write the relevant dynamics in standard notation and define every symbol.
  4. 4Apply no-arbitrage. In HJM, find the drift from the volatility. In a numerical question, use P(t,T) and forward rate relationships.
  5. 5Do the calculation carefully, with consistent units for time and rates.
  6. 6Interpret the result: what it says about the yield curve, correlation between maturities, or pricing.
  7. 7Comment on limitations, such as calibration difficulty, number of parameters, or non-Markov behaviour.

Quickest way: Quick comparison and forward-rate check

When to use it: Use for MCQs and short written parts that ask you to compare models or convert between bond prices and forward rates.

  1. For forward rates from bond prices, use the ratio P(t,T) ÷ P(t,T+δ) first.
  2. Subtract 1 and divide by δ for a simple forward rate.
  3. For continuous rates over [T1, T2], use ln[P(0,T1) ÷ P(0,T2)] ÷ (T2 − T1).
  4. For model-choice questions, remember: HJM fits today's curve exactly, LIBOR market models match caps via Black, two-factor short rate models give richer curve shapes but need calibration.
  5. For HJM drift questions, think: volatility in, drift out.

Common mistakes in Multifactor and Forward Rate Models

  • Treating the HJM drift as a free parameter to be chosen.

    Students are used to short rate models where the drift is a model input.

    Fix: Under the risk-neutral measure the HJM drift is fixed by the volatility through no-arbitrage. Only the volatility structure is chosen.

  • Saying more factors always give a better model.

    It sounds natural that a richer model is better.

    Fix: More factors add realism but also parameters, calibration difficulty and computation time. State both sides.

  • Confusing the short rate with the forward rate.

    Both are instantaneous rates and the notation looks similar.

    Fix: r(t) = f(t,t) only. For other T, f(t,T) is a forward rate and is not the expected future short rate in general.

  • Using annual compounding in a continuous formula, or the reverse.

    Forward LIBOR is simple-compounded, while f(t,T) is continuously compounded.

    Fix: Check the compounding convention before every calculation and say which one you use.

  • Claiming a one-factor model gives uncorrelated yields across maturities.

    Students mix up one factor with no dependence.

    Fix: In a one-factor model all yield changes are driven by the same shock, so they are perfectly correlated in the instantaneous sense. That is the weakness.

Worked examples

Example 1

Zero-coupon bond prices today are P(0,1) = 0.9500 and P(0,2) = 0.8900 (per 1 payable). Calculate the simple forward rate for the year from time 1 to time 2, and the continuously compounded forward rate for the same period.

Show the solution
  1. Accrual period δ = 1 year.
  2. Simple forward rate = [P(0,1) ÷ P(0,2) − 1] ÷ 1.
  3. P(0,1) ÷ P(0,2) = 0.9500 ÷ 0.8900 = 1.067416.
  4. Simple forward rate = 0.067416, or 6.74%.
  5. Continuous forward rate = ln(1.067416) ÷ 1.
  6. ln(1.067416) = 0.065240, or 6.52%.

Answer: Simple forward rate about 6.74% a year. Continuously compounded forward rate about 6.52% a year.

Example 2

Explain two advantages of a two-factor interest rate model over a one-factor short rate model, and state one disadvantage.

Show the solution
  1. A one-factor model has one shock, so all yield changes are perfectly correlated instantaneously. A two-factor model allows imperfect correlation between yields of different maturities.
  2. A two-factor model can generate more realistic yield curve movements, such as changes in slope and curvature as well as level shifts.
  3. It can also fit the volatility structure of different maturities better, which helps when pricing products sensitive to the shape of the curve.
  4. Disadvantage: it has more parameters, so calibration is harder and less stable, and computation is slower.

Answer: Advantages: imperfectly correlated yields across maturities, and more realistic curve shapes and movements. Disadvantage: more parameters, with harder calibration and more computation.

Exam tips

  • Be ready to explain in words why one factor is not enough. This is the most commonly asked point.
  • For HJM, state clearly: input is the initial forward curve and the volatility, and the drift follows from no-arbitrage.
  • Compare models in a short list: what is modelled, fit to today's curve, ease of calibration, typical use.
  • In numerical parts, write the compounding convention next to each rate.
  • For LIBOR market models, mention that they link naturally to Black pricing of caps and usually need Monte Carlo simulation.

Practice questions from Models of the term structure of interest rates

Multifactor and Forward Rate Models in other exams

The same ground in other exams, if you are preparing for more than one or want another angle on it.

Multifactor and Forward Rate Models: frequently asked questions

What is the Heath-Jarrow-Morton framework in simple terms?

It models how the whole forward rate curve moves over time. You choose the volatility of forward rates, and no-arbitrage then fixes the drift. The model fits today's market curve exactly because that curve is the starting point.

Why use a two-factor interest rate model?

A single factor makes all rates move together, which does not match real yield curves. A second factor lets short and long rates move differently. This gives more realistic changes in slope and shape.

What is a LIBOR market model?

It models a set of observable forward LIBOR rates, each over a fixed accrual period. Each rate is typically lognormal under its own forward measure. This is consistent with the Black formula for caplets, so it is popular for pricing.

Are HJM models Markov?

Not in general. The forward curve as a whole is Markov, but the short rate alone may not be, because it can depend on the path of past shocks. Special volatility choices make the short rate Markov.