1

The Binomial Model 1 In this lecture. . .

G G G G G

a simple model for an asset price random walk delta hedging no arbitrage the basics of the binomial method for valuing options risk neutrality

2 Introduction
We have seen a model for equities and other assets that is based on the mathematical theory of stochastic calculus. There is another, equally popular, approach that leads to the same partial differential equation, the BlackScholes equation, in a way that some people find more accessible, it can be made equally rigorous. This approach, via the binomial model for equities, is the subject of this lecture.

3 Equities can go down as well as up

In the binomial model we assume that the asset, which initially has the value 6 , can, during a timestep s W, either rise to a value X 6 or fall to a value Y 6 , with   Y    X . The probability of a rise is S and so the probability of a fall is  > S .
uS

vS
t

1.One timestep in a binomial random walk.

4
G

The three constants X , Y and S are chosen to give the binomial walk the same drift and standard deviation as that given by the stochastic differential equation model. Having only these two equations for the three parameters gives us one degree of freedom in this choice. This degree of freedom is often used to give the random walk the further property that after an up and a down movement (or a down followed by an up) the asset returns to its starting value, 6 . This gives us the requirement that
XY 

(1)

5 For the random walk to have the correct drift we need

S X   > S Y H
{ sW

Rearranging this equation we get

S  H { sW > Y X > Y 

(2)

Then for the random walk to have the correct standard deviation we need    {   s W S X   > 5 ;  (3) 0 Equations (1), (2) and (3) can be solved to give
X 
 

> { sW

 0

{  s W

L 6? 0 
 

> { sW

 0

{   s W

@>




4 The binomial tree

The binomial model allows the stock to move up or down a prescribed amount over the next timestep. If the stock starts out with value 6 then it will take either the value X 6 or Y 6 after the next timestep. We can extend the random walk to the next timestep. After two timesteps the asset will be at either X  6 , if there were two up moves,  X Y 6 , if an up was followed by a down or vice versa, or Y 6 , if there were two consecutive down moves. After three timesteps the asset can be at X  6 , X  Y 6 , etc.

7 The resulting structure is called the binomial tree.

u 4S

u 3 vS

u 2v 2S uv 3 S v 4S

2.The binomial tree.

8
u 4S

u 3 vS

u 2v2S

uv 3 S

v 4S

3.Binomial tree: schematic.

9 The top and bottom branches of the tree at expiry can only be reached by one path each, either all up or all down moves. Whereas there will be several paths possible for each of the intermediate values at expiry.
G G

Therefore the intermediate values are more likely to be reached than the end values. The binomial tree contains within it an approximation to the probability density function for the lognormal random walk.

10

5 An equation for the value of an option

G

Suppose, for the moment, that we know the value of the option at the time W  s W. For example, this time may be the expiry of the option. Now construct a portfolio at time W consisting of one option and a short position in a quantity d of the underlying. At time W this portfolio has value
h 9 > d 6

where the value 9 is to be determined.

11 At time W  s W the portfolio takes one of two values, depending on whether the asset rises or falls. These two values are
9


> @ :$

and

'

>

> @ Y6 

Since we assume that we know 9  , ' > , : , ; , $ and @ , the values of both of these expressions are known, and, in particular, depend on @ . Having the freedom to choose @ , we can make the value of this portfolio the same whether the asset rises or falls.

12 This is ensured if we make

'


> @ X6  9

>

> @ Y6 

This gives us the choice

@  9


> '

>

X > Y 6

(4)

when the new portfolio value is

h  sh 9


>

: '

> '

>

: > ; 

'

>

>

; '

> '

>

X > Y 

13 Since the value of the portfolio has been guaranteed, we can say that its value must coincide with the value of the original portfolio plus any interest earned at the risk-free rate; this is the no-arbitrage argument. Thus
sh U h s W

After some manipulation this equation becomes

9 9


> '

>

: > ;

:'

>

> ;'

  7 I 9: > Y 

(5)

This, then, is an equation for 9 given 9  , and 9 > , the option values at the next timestep, and the parameters : and ; describing the random walk of the asset.

14 To
I 9

we can write (5) as

0
7 I9

'  5 '

  > 5 '

>

(6)

where
S

H 7 I9 > Y X > Y

(7)

Interpreting S as a probability, this is just risk neutrality again.

G

And (6) is the statement that the option value at time 9 is the present value of the risk-neutral expected value at any later time. Supposing that we know '  and ' > we can use (6) to find ' . But do we know '  and ' > ?

15

6 Valuing back down the tree

We certainly know '  and ' > at expiry, time % , because we know the option value as a function of the asset then, this is the payoff function. If we know the value of the option at expiry we can find the option value at the time % > I 9 for all values of $ on the tree. But knowing these values means that we can find the option values one step further back in time.
G

Thus we work our way back down the tree until we get to the root. This root is the current time and asset value, and thus we find the option value today.

17 We are valuing a European call option, strike price 100 and expiry in four months. Todays asset price is 100, the volatility is 20%. The interest rate is zero. We use a timestep of one month so that there are four steps until expiry. Using these numbers we have I9          , X      , Y      and S      . As an example, after one timestep the asset takes either the value    @           or    @         . Working back from expiry, the option value at the timestep before expiry when 6      is given by
H
>   @    

     @       >       @     

    

18 Working right back down the tree to the present time, the option value when the asset is 100 is 6.14. Compare this with the theoretical, continuous-time solution (given by the BlackScholes call value) of 6.35. The difference is entirely due to the size and number of the timesteps. The larger the number of timesteps, the greater the accuracy.

19
6.36 6.34 6.32 6.3 6.28 6.26 6.24 6.22 6.2 6.18 6.16 6.14 0 50 100 150 200

4.Option price as a function of number of timesteps.

20

7 The greeks
The greeks are defined as derivatives of the option value with respect to various variables and parameters.
G

It is important to distinguish whether the differentiation is with respect to a variable or a parameter. If the differentiation is only with respect to the asset price and/or time then there is sufficient information in our binomial tree to estimate the derivative. The options delta, gamma and theta can all be estimated from the tree. On the other hand, if you want to examine the sensitivity of the option with respect to one of the parameters, then you must perform another binomial calculation. This applies to the options vega and rho for example. Let me take these two cases in turn.

21 From the binomial model the options delta is defined by

9


> '

>

X > Y 6

We can calculate this quantity directly from the tree, using the option value at the two points marked D, together with todays asset price and the parameters X and Y .
G D

5.Calculating the Greeks.

22 Estimating the other type of greeks, the ones involving differentiation with respect to parameters, is slightly harder. They are harder to calculate in the sense that you must perform a second binomial calculation. The vega is the sensitivity of the option value to the volatility
#9 # 

Suppose we want to find the option value and vega when the volatility is 20%. The most efficient way to do this is to calculate the option price twice, using a binomial tree, with two different values of . Calculate the option value using a volatility of D t , for a small number t , call the values you find 9 D . The vega is approximated by
'


> '>
t

23

8 Early exercise
American-style exercise is easy to implement in a binomial setting. The algorithm is identical to that for European exercise with one exception. We use the same binomial tree, with the same X , Y and S , but there is a slight difference in the formula for 9 . We must ensure that there are no arbitrage opportunities at any of the nodes. Q to mean the asset price at the Q th timestep, Introduce the notation 6 M at the node M from the bottom, T M T Q . In our lognormal world we have
6
Q M

 6 X Y

Q> M

where 6 is the current asset price. Also introduce 9 MQ as the option value at the same node. Our ultimate goal is to find 9 knowing the payoff, i.e. knowing 9 M1 for all T M T 0 where 0 is the number of timesteps.

24 In the American option problem, arbitrage occurs if the option value goes below the payoff at any time. If our theoretical value falls below the payoff then it is time to exercise. If we do then exercise the option its value and the payoff must be the same. If we find that
9 M
3  

> 'M

Q 

: > ;

 0

> 7 I9

:'M

3 

> ; ' M

Q  

: > ;

3 Payoff$ M 

then we use this as our new value.

25 But if
' M
Q   3  3  M Q  

> 'M

: > ;

 0

> 7 I9

:'

> Y 9 M

X > Y

3 Payoff6 M 

we should exercise, giving us a better value of

9
Q M Q Payoff6 M 

26 We can put these two together to get

U

Q M

P D[

'

3  M 

> '

Q  M

: > ;

 0

> 7 I9

:'

3  M

> Y9

Q  M 


 Payoff$
3 M

X > Y

This ensures that there are no arbitrage opportunities. This modification is easy to code.

27

9 The continuous-time limit

Equation (5) and the BlackScholes equation are more closely related than they may at first seem. Recalling that the BlackScholes equation is in continuous time, we examine (5) as s W  . First of all, we have that
: z   P

I9 

  P I9 

 ccc

and
; z  > P

I9 

  P I9 

 ??? 

Next we write
9  9 6 W 9


9 X 6 W  s W

and

>

 ' Y 6  W  s W

28 Expanding these expressions in Taylor series for small s W and substituting into (4) we find that
d z #9 #6

as

sW 

Thus the binomial delta becomes, in the limit, the BlackScholes delta. Similarly, we can substitute the expressions for 9 , 9  and ' > into (5) to find
' 9      P $ 


'


$

 7

' $

> U9  

This is the BlackScholes equation. Again, the drift rate { has disappeared from the equation.