Yield Curve Construction Using Short Rate

Models
Pablo Marchesi
January 2025

Abstract
This paper aims to provide a straightforward and accessible introduction to modelling the
yield curve, a key component in valuing interest rate derivatives. We utilize one-factor
short-rate models, which allow the construction of the entire yield curve by simulating the
interest rate for short-term borrowing. While this paper does not address the calibration
of these models, it includes examples of simulated yield curves for illustrative purposes.

Introduction
The valuation of interest rate derivatives is more complex than equity or FOREX deriva-
tives. This is mainly because the behaviour of the underlying, which in this case is the
interest rates, is more difficult to model compared to a stock or an exchange rate. Ad-
ditionally, the valuation of some products depends not only on one but several interest
rates with different maturities; in other words, these products depend on the entire yield
curve. As we will see later, modelling the yield curve is a challenging task, as this curve
can present very diverse shapes, and the volatility on each point of the curve is different.

Our approach to modelling the yield curve involves several steps. First, we simulate
the instantaneous short rate, which represents the risk-free rate applicable to an infinites-
imally short time period. Next, we compute the bank account process, representing the
value of one unit of currency invested in the money market and continuously reinvested
at the short rate. Using the bank account process, we then calculate the prices of zero-
coupon bonds for various maturities. Finally, we construct the yield curve by extracting
yields from these zero-coupon bond prices, creating the zero-coupon yield curve.

The key component in this methodology is selecting an appropriate short-rate model

and calibrating its parameters to fit current market prices. We will focus on several
aspects, such as whether the model can produce negative interest rates, the distribution of
the short rate, how suited the model is for Monte Carlo simulation, or if the model presents
a mean-reverting behaviour. We will address these issues in the following sections.

1
Yield Curve Construction Using Short Rate Models

Constructing the Yield Curve

We will start by defining the bank account process, B(t), which represents a risk-less
investment where profit is accrued continuously at the short rate (or risk-free) rate R(t).
Both the bank account and the short rate are stochastic processes. We will define R(t)
in the following section as it depends on our chosen model. We now give a more formal
definition of the bank account process [1]:

Definition 1 Bank account process: We define B(t) to be the value of a bank account
at time t ≥ 0. We assume B(0) = 1 and that the bank account evolves according to the
following differential equation:

dB(t) = R(t)B(t)dt, B(0) = 1 (1)

Where R(t) is the is the instantaneous rate at which the bank account accrues or the
short rate. As a consequence,
Z t 
B(t) = exp R(s)ds (2)
0

Figure 1: Bank Account Process

Additionally, we can compute the discount process, D(t), from the bank account
process as follows:
 Z t 
1
D(t) = = exp − R(s)ds (3)
B(t) 0

This is also called the stochastic discount factor. When dealing with equities and
exchange rates, the bank account and the discount factor are deterministic as R(t) is
assumed to be constant. However, in our case, we are forced to drop the deterministic
setup as we will have to model R(t) as a stochastic process.

The next step will be computing the price of zero coupon bonds from the discount
process. The zero coupon bonds represent the building blocks of the yield curve, as they

2
Yield Curve Construction Using Short Rate Models

provide a direct measure of spot rates in the market. These bonds pay a certain face
amount, which we take to be 1, at a fixed maturity date of T . More formally, we can
define a zero coupon bond as [1]:

Definition 2 Zero-coupon bond: A T -maturity zero-coupon bond (pure discount bond)

is a contract that guarantees its holder the payment of one unit of currency at time T ,
with no intermediate payments. The contract value at time t < T is denoted by P (t, T ).
Clearly, P (T, T ) = 1 for all T .

The price of a zero-coupon bond can be calculated by using the risk-neutral pricing
formula, which states that the discounted price of the bond must be a martingale under
the risk-neutral measure. Therefore, the price of a zero-coupon bond at time t = 0 should
satisfy:
  Z T 
P (0, T ) = Ẽ[D(T ) P (T, T )] = Ẽ exp − R(s)ds (4)
0

We can generalize this expression for any time t such that 0 ≤ t ≤ T , getting the
following:
  Z T 
P (t, T ) = Ẽ exp − R(s)ds (5)
t

Intuitively, the price of a zero-coupon bond is the average discounted value across all
possible future interest rate paths over the bond’s life (until T ) at any time t. We can
also express the price of a zero coupon bond as:

P (t, T ) = e−Y (t,T )(T −t) (6)

Where Y (t, T ) is the constant rate of continuously compounding interest rate between
times t and T , or, in short, the yield between t and T . From the previous expression, we
can deduce the following:
1
Y (t, T ) = − log P (t, T ) (7)
T −t
This, in fact, is the function of the yield curve (also known as the zero-cupon curve),
as we can compute the yield at any time t for several bond maturities T . For example,
the 30-year rate at time t is Y (t, t + 30). This is an example of a long rate. If we make
T approach t, in other words, if we take a very short maturity bond, we get:

lim Y (t, T ) = R(t) (8)

T →t+

This implies that the beginning of the yield curve is the short rate R(t), and the fol-
lowing points of the curve are then calculated with (7). Note that once we adopt a model
for the short rate R(t), we can determine the long rates Y (t, T ). We have accomplished
our goal of constructing a yield curve using an arbitrary short-rate model.

Our next step will be choosing an appropriate model for R(t), which we will discuss
in the following section.

3
Yield Curve Construction Using Short Rate Models

Figure 2: Yield curve constructed with a Vasicek model

Short-Rate Models
A short-rate model can be described with the following stochastic differential equation:

dR(t) = µ (t, R(t)) dt + σ (t, R(t)) dW̃ (t) (9)

where W̃ (t) is a Brownian Motion under the risk-neutral probability measure P̃ and
µ and σ are known functions representing the drift and the diffusion coefficients respec-
tively. In the equity and FOREX world, we assume an evolution of the price of the
underlying under the actual measure, and then we switch to a risk-neutral measure once
we have determined the market price of risk. However, interest rates are not the price of
an asset. Thus, we can not compute the market price of risk, therefore we build these
models using a risk-neutral measure from the outset. For a more detailed explanation,
refer to Chapter 10 of [3].

We will focus on one-factor short-rate models, which rely on a single stochastic dif-
ferential equation to describe the behaviour of interest rates. In such models, all rates
move in the same direction over short time intervals, though not necessarily by the same
amount. This allows us to model parallel shifts in the yield curve. However, these models
cannot capture more complex changes in the yield curve, such as variations in its slope
or curvature over time.

For the task of choosing a suitable model, it is important to consider the following
key aspects:

• Does the model allow negative interests rates, R(t) < 0?

• What is the distribution of the short-rate? Is it a fat-tailed distribution?

• Are bond prices P (t, T ) computable from the dynamics of the model?

• How suitable is the model for Monte Carlo simulation?

• Is the model mean-reverting (as interest rates tend to behave in this way)?

4
Yield Curve Construction Using Short Rate Models

In the context of this paper, we will watch closely if the yield curve calculated from
the chosen model can fit the actual term structure of interest rates, in other words, if
Y (0, T ) can fit the current spot yield curve.

The model’s mean-reverting behaviour is a crucial aspect. Without incorporating this

dynamic for interest rates, the model may produce unrealistic outcomes. When interest
rates are high, the economy typically slows down, leading to reduced demand for funds
from borrowers, which causes rates to decrease. Conversely, when interest rates are low,
borrower demand for funds tends to increase, driving rates upward. We can confirm this
behaviour empirically by taking a look on the evolution of any US Treasury Yield.

Figure 3: Interest rates usually present a mean reverting behaviour

This paper focuses on three one-factor short-rate models: Vasicek, CIR, and Hull-
White. Although numerous other models exist, including multi-factor models (which are
more realistic), our analysis will be confined to these three models.

1. Vasicek Model
The Vasicek model is a one-factor short-rate model that assumes a mean-reverting be-
haviour, ensuring that interest rates do not drift to extreme values. The following stochas-
tic differential equation defines the model:

dR(t) = κ(θ − R(t))dt + σdW̃ (t) (10)

where:

• R(t) is the short rate at time t

• κ > 0 is the speed of mean reversion

• θ is the long-term mean level of the rate

• σ > 0 is the volatility

• W̃ (t) is a standard Brownian motion under the risk-neutral measure

5
Yield Curve Construction Using Short Rate Models

Above is a plot of the probability density function of R(T ) (at maturity) and several
yield curves constructed with the previously outlined methodology for different time in-
stants.

Figure 4: Vasicek model with parameters R(0) = 0.03, θ = 0.05, κ = 0.1 and σ = 0.01

Note that the PDF allows for negative values of the short rate. This is one major draw-
back of the Vasicek model. For some parameter values (especially if σ is large enough),
the model produces negative values for R(t). This is, in general, a not-so-desirable feature
as interest rates, under normal market conditions, should not be negative.

If we look at the yield curves constructed with the model, we can see that they vary
over time, although they tend to produce just parallel shifts, failing to reproduce changes
on the slope or the curvature. This means that the Vasicek model will fail to reproduce
some shapes of the yield curve no matter how the parameters are chosen, making the
model useless in some market scenarios.

2. CIR Model
The Cox-Ingersoll-Ross (CIR) model is a one-factor short-rate model that ensures mean-
reverting behaviour while preventing negative interest rates by using a square root diffu-
sion process. The following stochastic differential equation defines the model:
p
dR(t) = κ(θ − R(t))dt + σ R(t)dW̃ (t) (11)
which is very similar to (10), but incorporates the square root term in the diffusion
part of the equation. This fact makes the model more realistic as the CIR model will not
produce negative values for R(t) (if the parameters don not take extreme values). We
can confirm this by observing the PDF of the model (as shown below).

Like the Vasicek model, the CIR model cannot reproduce certain yield curve shapes.
This can be improved by using exogenous term structure models, which incorporate the
current spot yield curve into the model by adding a time-varying parameter so the model
can reproduce any yield curve shape at t = 0 and allows for more realistic modelling of

6
Yield Curve Construction Using Short Rate Models

the yield curve for other time instants. In the next section, we study the Hull-White
model, which is an example of an exogenous model.

Figure 5: CIR model with parameters R(0) = 0.03, θ = 0.05, κ = 0.1 and σ = 0.01

3. Hull-White Model
The Hull-White model is an extension of the Vasicek model that allows for time-dependent
mean reversion and volatility parameters, making it more flexible for fitting the initial
term structure of interest rates. It is defined by the following stochastic differential
equation:

dR(t) = κ(θ(t) − R(t))dt + σdW̃ (t) (12)

Note that the long-term mean parameter θ now is time-dependent, making the model
more flexible and able to produce more complex shapes of the yield curve. θ(t) must be
chosen so as to exactly fit the term structure of interest rates currently observed in the
market. The analytic expression for this parameter is the following:

∂f M (0, t) σ2
+ κf M (0, t) + 1 − e−2κt


θ(t) = (13)
∂T 2κ
Where f M (0, T ) is the market instantaneous forward rate at time 0 for the maturity
T , which can be computed from the actual spot yield curve observed in the market,
Y M (0, T ):

∂Y M (0, T )
f M (0, T ) = Y M (0, T ) + T (14)
∂T
For a more detailed explanation, refer to [1] chapter 3. One of the drawbacks of the
Hull-White model is that it allows negative interest rates (as can be seen in the PDF
below). Note that despite computing the expression of θ(t), we are not able to reproduce
the current spot curve as we would need to calibrate the model for the other parameters
κ and σ.

7
Yield Curve Construction Using Short Rate Models

Figure 6: Hull-White model with parameters R(0) = 0.04, θ = θ(t), κ = 0.05 and
σ = 0.03

Implementation
In this section, we will briefly comment on the implementation of the models and the
computation of the yield curves. We have taken a Monte Carlo approach for the short-
rate calculation by discretizing the stochastic differential equations and calculating the
values for R(t) in an iterative way. For example, in the case of the Vasicek model, the
equation (10) becomes:
√
Rt+∆t = Rt + κ(θ − Rt )∆t + σ ∆tZ (15)
where Z ∼ N (0, 1) is a standard normal random variable. Regarding the yield curve,
once the short rate is calculated, we have made use of expressions (5) and (7) in order to
come up with the values of Y (t, T ).

The PDFs of the short-rate models have been calculated empirically using a Gaussian
kernel density estimator, which has the following expression:
n
(x − xi )2
 
1 X
fˆ(x) = √ exp − (16)
nh 2π i=1 2h2
where:

• n is the number of data points,

• h is the bandwidth (smoothing parameter),

• xi are the observed data points.

For the Hull-White model, we have computed θ(t) using expressions (13) and (14)
and inputting the current spot yield curve observed from the market.

For a more detailed explanation of the implementation in Python, please click here.

8
Yield Curve Construction Using Short Rate Models

Figure 7: Current spot and forward curves (at t = 0)

Conclusion
In conclusion, we successfully developed a model to construct interest rate curves from
an arbitrary short-rate model. However, the models we studied fall short in capturing
the complexities of the yield curve. Some produce unrealistic scenarios, such as negative
interest rates, while others fail to fully explain the curve’s behaviour.

To address these shortcomings, future work could explore alternative short-rate mod-
els, particularly exogenous models with time-varying coefficients or multi-factor models
that better capture the dynamics of the yield curve.

References
[1] Damiano Brigo, Fabio Mercurio, et al. Interest rate models: theory and practice,
volume 2. Springer, 2001.

[2] John C Hull and Sankarshan Basu. Options, futures, and other derivatives. Pearson
Education India, 2016.

[3] Steven E Shreve et al. Stochastic calculus for finance II: Continuous-time models,
volume 11. Springer, 2004.