Scribd
Upload
40 views4 pages

Interest Rate Models 19.1 Basic Elements in A Bond Market: T T T U T

The document summarizes key elements of interest rate models, including three classical models - Vasicek, Cox-Ingersoll-Ross (CIR), and Heath-Jarrow-Morton (HJM). [1] The Vasicek and CIR models describe the stochastic process of the short-term interest rate, while HJM models the term structure directly. [2] The CIR model is an improvement over Vasicek since it ensures the short rate never becomes negative. [3] HJM provides greater flexibility than Vasicek and CIR but requires more complex computation.

Uploaded by

owltbig
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
40 views4 pages

Interest Rate Models 19.1 Basic Elements in A Bond Market: T T T U T

The document summarizes key elements of interest rate models, including three classical models - Vasicek, Cox-Ingersoll-Ross (CIR), and Heath-Jarrow-Morton (HJM). [1] The Vasicek and CIR models describe the stochastic process of the short-term interest rate, while HJM models the term structure directly. [2] The CIR model is an improvement over Vasicek since it ensures the short rate never becomes negative. [3] HJM provides greater flexibility than Vasicek and CIR but requires more complex computation.

Uploaded by

owltbig
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Lecture 19 Interest Rate Models

19.1 Basic elements in a bond market


R
Consider a market money account B = {Bt } with B0 = 1, Bt = exp( 0t ru du), t ∈ [0, T ], where
the short rate {rt } is assumed to be a stochastic process. The bank account B and related term
structures ({B(t, τ )}, or {Y (t, τ )}, or {f (t, τ )}) in a bond market serve as underlying assets similar
to stocks in a stock market. We will introduce several interest rate models and use them to price
various derivatives defined on the underlying assets. Note that each fixed-income asset or derivative
may have its own maturity time.

Let (Ω, F, P ) be a probability


³R space´equipped with a filtration {Ft }t∈[0,T ] . Assume {rt } is an
T
adapted process with P 0 |rt | dt < ∞ = 1. The filtration {Ft }t∈[0,T ] is usually generated by a
multidimensional Brownian motion (called shocks or factors) as a source of uncertainty. Since our
main purpose is derivative pricing, we assume the existence of a risk neutral measure Q, equivalent
to P following the Girsanov transformation (detail skipped), under which the discounted processes
of all underlying assets are martingales. This will rule out arbitrage opportunities, according to
the (continuous-time) Fundamental Theorem of Asset Pricing, which we have not presented. For
0 ≤ t < τ ≤ T , three equivalent term structures can be defined as follows:

• Zero-coupon bond: Let B(t, τ ) be the time-t price of a bond with maturity τ (called τ -bond)
and the par value B(τ, τ ) = 1. The risk-neutral valuation principle implies
· µ Z τ ¶¯ ¸
¯
B(t, τ ) = EQ exp − ru du ¯¯ Ft .
t

• Yield Y (t, τ ), defined by

−1
Y (t, τ ) = log B(t, τ ),
τ −t
which follows from

B(t, τ ) exp[Y (t, τ )(τ − t)] = 1,

i.e. Y (t, τ ) is the constant compounding interest rate that brings B(t, τ ) to the par value.

• Forward rate f (t, τ ), defined by

∂
f (t, τ ) = − log B(t, τ ),
∂τ
which follows from
·Z τ ¸
B(t, τ ) exp f (t, u)du = 1,
t

1
i.e. f (t, τ ) is the instantaneous rate locked in at t for risk-free borrowing or lending at τ , with
f (t, t) = rt . Intuitively, f (t, τ ) should be interpreted as the interest rate over the infinitesimal
time interval [τ, τ + ∆τ ] as seen at time t.

To ease the notation, assume 0 ≤ τ ≤ T and consider a European option g(B(τ, T )) with
maturity τ written on a zero-coupon bond with maturity equal to the horizon T . The option g
can be hedged by a dynamic portfolio h = {(h0t , h1t ), t ∈ [0, T ]}, where h0t Bt and h1t B(t, T )
represent the time-t balance of the bank account and the time-t holding value of the zero-coupon
bond respectively. The time-t value of the portfolio is

Vt = h0t Bt + h1t B(t, T ), (19.1)

with the corresponding self-financing condition

dVt = h0t dBt + h1t dB(t, T ). (19.2)

Under certain regularity conditions, the time-t value of the option g is given by
· µ Z τ ¶ ¯ ¸
¯
Vt = EQ exp − ru du g(B(τ, T ))¯¯ Ft , (19.3)
t

which can be calculated in each problem with a specified interest rate model.

19.2 Three classical models

We will present three continuous-time single factor interest rate models in this section, whose
discrete-time versions appeared in Lecture 8. In each model, the SDE represents the risk neutral
dynamics under Q.

19.2.1 Vasicek model

Assume the short rate under Q satisfies the SDE

drt = a(b − rt ) dt + σ dWt (19.4)

with positive constant parameters a, b and σ. Note that X = {Xt } with Xt = rt − b is an Ornstein-
Uhlenbeck (O-U) process satisfying the SDE

dXt = −aXt dt + σ dWt .

Hence X is a Gaussian process, which implies that Q(rt < 0) > 0 for every t ∈ [0, T ] — an
unreasonable behavior in practice. Pricing bond options would be easy due to nice properties of
the O-U process.

2
19.2.2 Cox-Ingersoll-Ross (CIR) model

The short rate process under Q follows

√
drt = a(b − rt ) dt + σ rt dWt (19.5)

with constant parameters a, b and σ. Suppose r0 = x > 0 and define

Tx,0 = inf{t > 0 : rt = 0}, or Tx,0 = ∞ if rt > 0 ∀ t > 0.

The following proposition shows that the CIR model is an important improvement of the Vasicek
model.

Proposition 19.1 There are three cases based on different ranges of the parameters:

Case 1 If ab ≥ σ 2 /2, then Q(Tx,0 = ∞) = 1 ∀ x > 0.

Case 2 If 0 ≤ ab < σ 2 /2 and a ≥ 0, then Q(Tx,0 < ∞) = 1 ∀ x > 0.

Case 3 If 0 ≤ ab < σ 2 /2 and a < 0, then 0 < Q(Tx,0 < ∞) < 1 ∀ x > 0.

Proposition 19.1 provides us with a guideline for setting the parameters a, b, σ. {rt } in (19.5) is an
example of Feller’s square-root diffusions which enjoy great popularity in financial economics. It is
a Markov process with a closed-form transition density: the conditional density of ru given rt for
t < u is a scaled version of non-central χ2 distribution. Prices of the zero-coupon bond and some
bond options under a CIR model can be obtained with explicit expressions.

19.2.3 Heath-Jarrow-Morton (HJM) model

The Vasicek and CIR models, although very simple, do not model term structures directly.
When market term structure data {B(t, τ ), τ ∈ [t, T ]} are observed at time t, they are inconsistent
with the output generated by these two short rate models. In contrast, HJM proposes a general
mechanism that models term structures directly and takes observed term structures as inputs.

df (t, τ ) = α(t, τ ) dt + σ(t, τ ) dWt (19.6)

given the initial term structure f (0, τ ) = g(τ ), τ ∈ [0, T ], where α(·), σ(·) satisfy the HJM drift
condition
Z τ
α(t, τ ) = σ(t, τ ) σ(t, u) du (19.7)
t

3
to avoid arbitrage.

The integral form of (19.6)


Z t Z t
f (t, τ ) = g(τ ) + α(u, τ ) du + σ(u, τ ) dWu , (19.8)
0 0

along with f (t, t) = rt , will give us


Z t Z t
rt = f (0, t) + α(u, t) du + σ(u, t) dWu . (19.9)
0 0

Note that in a general HJM model, the drift coefficient α(t, τ ) and volatility coefficient σ(t, τ ) are
not just functions of “current state” f (t, τ ), they also involve the history {f (u, τ ), 0 < u < t}. For
fixed τ , the process {f (t, τ ), 0 ≤ t ≤ τ } is non-Markovian. Therefore, the computation required
for the HJM model is extremely complicated, as a price to pay for greater modeling flexibility than
the Vasicek and CIR models.

Multi-factor versions of Vasicek, CIR and HJM are available that further improve the results
in empirical studies.

You might also like