MScFE 622 Continuous-time Stochastic Processes – Video Transcripts Module 7

Video Transcripts
Module 7
MScFE 622
Continuous-time Stochastic
Processes

© 2019 - WorldQuant University – All rights reserved.

Revised: 07/07/2020
1
MScFE 622 Continuous-time Stochastic Processes – Video Transcripts Module 7

Table of Contents

Unit 1: Term Structure ...................................................................... 3

Unit 2: Short Rate Models ................................................................ 5
Unit 3: Other Short Rate Models ..................................................... 7
Concluding Video............................................................................... 9

© 2019 - WorldQuant University – All rights reserved.

2
MScFE 622 Continuous-time Stochastic Processes – (Lecture 1) Module 7: Unit 1

Unit 1: Term Structure

Hi, in this video we introduce stochastic interest rates in continuous time.

So, all along, we have assumed that 𝑟, the interest rate, is constant for all maturities. This then
implies, of course, that the bank account evolves according to the following SDE:

𝑑𝐵! = 𝑟𝐵! 𝑑𝑡, 𝐵" = 1

Solving this SDE gives us:

𝐵! = 𝑒 #! , 𝐷! = 𝑒 $#!

where 𝐷! is the discount factor.

Now, we are going to assume that the interest rate, 𝑟, is not constant, and, therefore, the bank
account evolves according to the following SDE:

𝑑𝐵! = 𝑟! 𝐵! 𝑑𝑡, 𝐵" = 1

If we solve this SDE, we get:

"
𝐵! = 𝑒 ∫# #! &'

So, instead of just simply multiplying 𝑟 by 𝑡 when it was constant, we now have to evaluate an
"
integral as we do above. Of course, the discount factor, 𝐷! , is equal to 𝑒 $ ∫# #! &' .

We are also going to generalize the discount factor and define 𝐷(𝑡, 𝑇) as the discount factor
" $
between periods 𝑡 and 𝑇. Using the formula 𝐷! = 𝑒 $ ∫# #! &' , we get 𝑒 $ ∫" #! &'
.

This interest rate, 𝑟! , that we are going to be using now is called a short rate and in the next video
we will see how to model it as a stochastic process.

We also introduce the zero-coupon bond, which is an asset that has the following payoff:

A zero-coupon bond that expires at time 𝑇 is an asset that pays one unit of currency at time 𝑇.

© 2019 - WorldQuant University – All rights reserved.

3
MScFE 622 Continuous-time Stochastic Processes – (Lecture 1) Module 7: Unit 1

What we are interested in, however, is the price of that zero-coupon bond at an intermediate time,
𝑡, between 0 and 𝑇. We will denote this price by 𝑃(𝑡, 𝑇). So, this is the price of a zero-coupon bond
with maturity, or, as it is sometimes called, expiry time, 𝑇. We are only evaluating it at time 𝑡.

Now, if ℙ∗ is an ELMM, then, since the discounted prices must be martingales, we have the
following formula, where the discounted zero-coupon bond price at time 𝑡 must be equal to the
conditional expectation of the dicounted zero-coupon bond price at time 𝑇, meaning that we have
to discount it by 𝑇, given ℱ𝓉 :

𝐷! 𝑃(𝑡, 𝑇) = 𝐸 ∗ (𝑃(𝑇, 𝑇)𝐷* |ℱ𝓉 )

That is the risk-neutral pricing formula that we have been using all along and we are now applying
it to zero-coupon bond prices.

Now, since the zero-coupon bond at time 𝑇 pays $1, (which means that its value is equal to 1), we
get 𝐸 ∗ (𝐷* |ℱ! ), and, if we rearrange terms, we get the pricing formula for bonds, which says that
𝑃(𝑡, 𝑇), the price of a zero-coupon bond at time 𝑡 that has maturity time 𝑇, is simply equal to the
expected value under the pricing measure, or the ELMM we are using. We then divide by 𝐷! and,
since 𝐷* is adapted, we can take it inside and write it as:

𝐷
𝑃(𝑡, 𝑇) = 𝐸 ∗ 7 𝐷* 8ℱ! 9
!

So, this is the bond-pricing formula when we have stochastic interest rates.

Now that we have introduced stochastic interest rates in continuous time, in the next video we are
going to look at a short rate model called the Vasicek model.

© 2019 - WorldQuant University – All rights reserved.

4
MScFE 622 Continuous-time Stochastic Processes – (Lecture 2) Module 7: Unit 2

Unit 2: Short Rate Models

The Vasicek Model

Hi, in this video we introduce a short rate model known as the Vasicek model.

In the Vasicek model, the short rate, 𝑟, satisfies the following SDE:

𝑑𝑟! = 𝛼(𝜃 − 𝑟! ) 𝑑𝑡 + 𝜎𝑑𝑊!

where α, θ and 𝜎 are all positive constants. So, it is mean-reverting in this form. This is actually
called an Ornstein-Uhlenbeck process, which you should remember from the second module.

To solve this SDE, we apply Ito's Lemma to 𝑟! times 𝑒 +! . If we find the find the stochastic
differential of 𝑑(𝑟! 𝑒 +! ), we get 𝑒 +! 𝑑𝑟! , by taking the derivative with respect to 𝑟! , plus, taking this
part here, we get 𝛼𝑒 +! 𝑟! 𝑑𝑡. Written in full:

𝑑(𝑟! 𝑒 ,! ) = 𝑒 ,! 𝑑𝑟! + 𝛼𝑒 ,! 𝑟! 𝑑𝑡

We can then simplify this to get:

(𝛼𝑒 ,! 𝜃 − 𝛼𝑒 ,! 𝑟! )𝑑𝑡 + 𝑒 ,! 𝜎𝑑𝑊! + 𝛼𝑒 ,! 𝑟! 𝑑𝑡

We can further simplify this, when we realize that we can cancel out two parts of the equation
(−𝛼𝑒 +! 𝑟! and 𝛼𝑒 +! 𝑟! 𝑑𝑡), to get:

α𝑒 +! θ𝑑𝑡 + σ𝑒 +! 𝑑𝑊!

So, this is the stochastic differential, and it doesn't depend on any unknown processes so we can
integrate it to solve for 𝑟! .

Doing this, on the left-hand side, we get:

! !
𝑟! 𝑒 +! − 𝑟" = G α𝑒 +' θ𝑑𝑠 + G σ𝑒 +' 𝑑𝑊'
" "

© 2019 - WorldQuant University – All rights reserved.

5
MScFE 622 Continuous-time Stochastic Processes – (Lecture 2) Module 7: Unit 2

So, this is the solution for 𝑟! , which we can further simplify to get:

! !
𝑟! = 𝑟" 𝑒 $+! + 𝑒 $+! G α𝑒 +' 𝜎𝑑𝑠 + 𝑒 $+! G 𝜎𝑒 +' 𝑑𝑊'
" "

!
So, this is the solution for 𝑟! . As you can see, this part, 𝑟" 𝑒 $+! + 𝑒 $+! ∫" 𝛼𝑒 +' 𝜎𝑑𝑠, is deterministic,
meaning that it doesn't have any randomness in it. The only random part is the end of the
" %!
equation, 𝑒 $+! ∫# -. 𝑑𝑊' , and, since the integrand, 𝜎𝑒 +' , in this case is also deterministic, it follows
that 𝑟! has a normal distribution.

[When looking to calculate the mean of the equation, we look at the first part of the equation only,
as the mean of the second part of the equation is 0. If we evaluate the integral in the first part of
the equation, we get:

𝐸(𝑟! ) = 𝑟! 𝑒 $+! + θ(1 − 𝑒 $+! )

So, that is the mean of the first part of the equation. If we add it to the variance of the second part
of the equation, we get:

𝜎/
Var(𝑟! ) = (1 − 𝑒 $/+! )
2α

So, this is the mean and the variance of the normal distribution.

Of course, a huge disadvantage of this is that it implies that interest rate can in fact be negative
with a positive probability since the normal distribution does go negative.

Now that we have introduced the Vasicek model, in the next video we are going to look at other
interest rate models.

© 2019 - WorldQuant University – All rights reserved.

6
MScFE 622 Continuous-time Stochastic Processes – (Lecture 3) Module 7: Unit 3

Unit 3: Other Short Rate Models

Hi, in this video we discuss some other popular interest rate models.

The Ho-Lee model

This model states that the short rate, 𝑟, evolves according to the following SDE:

𝑑𝑟! = 𝜃! 𝑑𝑡 + 𝜎𝑑𝑊!

As you can see, 𝜃 is a function of time and 𝜎 is a positive constant.

The solution to the SDE is simple: it just says that 𝑟! has a normal distribution.

The Vasicek model

As a reminder, we looked at this model in the previous video. This model states that the short rate
evolves according to the following SDE:

𝑑𝑟! = 𝛼(𝜃 − 𝑟! )𝑑𝑡 + 𝜎𝑑𝑊!

A huge advantage of the Vasicek model, compared to the Ho-Lee Model, is that it has the property
of mean reversion, 𝛼(𝜃 − 𝑟! ). If 𝑟! is greater than 𝜃, then the drift becomes negative and it pulls the
process down towards 𝜃 and, if 𝑟! is less than 𝜃, the drift becomes positive and pulls it towards 𝜃.

We can look at it on a graph: if we have 𝜃 on the y-axis (which is sometimes called the level of
mean reversion) then, if 𝑟! is less than 𝜃, the drift is positive, meaning that this is pulled towards 𝜃;
and, as soon as 𝑟! is greater than 𝜃, the drift becomes negative and the interest rate 𝑟! is pulled
downwards towards 𝜃 itself.

However, 𝑟! still has a normal distribution, which implies that we can have negative interest rates
with positive probability.

© 2019 - WorldQuant University – All rights reserved.

7
MScFE 622 Continuous-time Stochastic Processes – (Lecture 3) Module 7: Unit 3

The Hull-White model

This model's SDE for 𝑟! is similar to the Vasicek model, except that the parameters are now time-
dependent rather than constant:

𝑑𝑟! = (𝜃(𝑡) − 𝛼(𝑡)𝑟! )𝑑𝑡 + 𝜎(𝑡)𝑑𝑊!

The advantage of this model is that the level of mean reversion, 𝜃, can be varied. So, instead of
having one level of mean reversion on our graph, we can have a level of mean reversion that
changes with time. This has huge advantages in real-life modeling because this model fits the
market data better than the Vasicek model.

The Cox-Ingersoll-Ross (CIR) model

In this model, the short rate, 𝑑𝑟! , evolves according to the following SDE:

𝑑𝑟! = 𝑎(𝑏 − 𝑟! )𝑑𝑡 + 𝜎O𝑟! 𝑑𝑊!

An important assumption that we make is that 2𝑎𝑏 ≥ 𝜎 / . This assumption ensures that the
interest rates that we get from this model are always positive, which is a huge advantage
compared to the previous three models because those three models can give us negative interest
rates with positive probability. With the CIR model, on the other hand, the interest rates are
always positive.

Now that we have covered these interest rate models we have reached the end of our final
module.

© 2019 - WorldQuant University – All rights reserved.

8
MScFE 622 Continuous-time Stochastic Processes – (Concluding Video) Module 7: Unit 4

Concluding Video
Congratulations on finishing Continuous-time Stochastic Processes, the fourth course in the
WorldQuant University Master of Science in Financial Engineering.

In this course, we explored stochastic processes in continuous time and applied that knowledge to
the pricing of derivatives in continuous-time asset price models.

In the next course, Computational Finance, you will be introduced to the Python programming
language, Monte Carlo methods, as well as applications to option pricing and risk measurement.

Now that you have completed the Continuous-time Stochastic Processes course, you should be
able to:

• Define and identify Brownian motion processes in multiple dimensions.

• Solve stochastic differential equations.
• Apply Ito's Lemma for continuous semimartingales.
• Apply Girsanov's Theorem to construct equivalent local martingale measures.
• Price and hedge derivatives in various asset price models.
• Derive the Black-Scholes partial differential equation.
• Construct asset price models based on Lévy processes.
• Price interest rate derivatives.

Thank you for your engagement throughout this course. We hope that you will continue to enjoy
the rest of the program.

Good luck!

© 2019 - WorldQuant University – All rights reserved.

9