Week 11.

Binomial Trees II
AF4317. Derivative Securities

Prof. Byoung Uk Kang

1/3/4 April 2025

1 / 30
Lecture Outline

• Risk-Neutral Valuation vs. DCF

• Two-Step Binomial Tree

• N-Step Binomial Tree

• Advanced Topics in Binomial Models

• American Options

• Determination of u and d

2 / 30
Risk-Neutral Valuation vs. DCF

• The option price from the risk-neutral valuation is the same as the
price from the DCF.

• If required return on stock α is higher than the risk-free rate r , it

follows that p < p ∗ .

• It implies that in the risk-neutral valuation, we amplify the probability

of a bad outcome for stock investors.

• We interpret this as the probability being modified to incorporate

investors’ risk-aversion.

3 / 30
Risk-Neutral Valuation

• Risk-neutral valuation is a probabilistic interpretation of the option

pricing formula obtained from the replicating portfolio.

• This interpretation is also useful for multi-step binomial models.

4 / 30
Two-Step Binomial Models

5 / 30
Two-Step Binomial Model
• The current stock price is S0 and may go up by u or down by d in a
time step. Each time step is ∆t and the risk-free interest rate is r per
annum.

• A European call option has the strike price of K and expires in two
steps. What is the option price?
S0 uu
fuu
S0 u u
u fu d
S0 S0 ud
f fud
d S0 d u
fd d
S0 dd
fdd
6 / 30
Two-Step Binomial Model

• We start at the option expiration date and find the option payoff at
each stock price then.

• At T = ∆t, each price and the following prices can be seen as

one-step binomial tree. Thus, we can use the pricing formula of
one-step models.

= e −r ∆t [pfuu + (1 − p)fud ]

fu
fd = e −r ∆t [pfud + (1 − p)fdd ]

e r ∆t −d
where p = u−d .

7 / 30
Two-Step Binomial Model

• At T = 0,

f0 = e −r ∆t [pfu + (1 − p)fd ]
h    i
= e −r ∆t p e −r ∆t [pfuu + (1 − p)fud ] + (1 − p) e −r ∆t [pfud + (1 − p)fdd ]
h i
= e −2r ∆t p 2 fuu + 2p(1 − p)fud + (1 − p)2 fdd

• This is consistent with the probabilistic interpretation of p.

• In risk-neutral valuation, p 2 , 2p(1 − p), and (1 − p)2 are probabilities
of reaching top, middle, and bottom final nodes.

8 / 30
Two-Step Binomial Model - Put Option
• To price a put option, we use put payoffs at the option expiration.
The rest of calculation is the same as the call valuation.

Q. Consider a 2-year European put with K =$52 on a stock with

S0 =$50. Suppose that a time step is 1 year, and in each time step
the stock price moves either up or down by 20%. The risk-free
interest rate is 5%. What is the price of the put option?

72
fuu
60 u
u fu d
50 48
f fud
d 40 u
fd d
32
fdd

9 / 30
N-Step Binomial Models

10 / 30
N-Step Binomial Model

• Suppose that there are N time steps until the option maturity and
each time step is ∆t.

• The risk-neutral probability of an increase in stock price during each

r ∆t −d
step is p(= e u−d ).

• There are N + 1 nodes at the expiration. Let node j denote the final
stock price when the price moves upward j times and downward N − j
times. There, the final stock price would be

S0 u j d N−j ,

where j = 0, 1, ..., N.

11 / 30
N-Step Binomial Model
• To determine the price of an European option, we need the
probability of reaching each node at the expiration.

• The probability of reaching the node j is

 
N
p j (1 − p)N−j
j

• There are multiple

 paths leading to the node j. The number of the
N
paths is , which is j-combinations from a set of N elements.
j
 
• How to calculate N ?
j
 
• In algebra, N N!
= j!(N−j)!
.
j
• In Excel, use “combin(N,j)”.

12 / 30
N-Step Binomial Model
• For each node, the probability and option payoff is as follows:
No. of up No. of down Probability Stock price at T Option payoff
0 N p 0 (1 − p)N S0 u 0 d N f0
 
N
1 N −1 p 1 (1 − p)N−1 S0 u 1 d N−1 f1
1
.. .. .. .. ..
. .   . . .
N
j N −j p (1 − p)N−j
j
S0 u j d N−j fj
j
.. .. .. .. ..
. . . . .
N 0 p (1 − p)0
N
S0 u N d 0 fN

• The price of European option is then

N  
−r (N∆t)
X N
e p j (1 − p)N−j fj
j
j=0

where fj is the option payoff at node j.

13 / 30
N-Step Binomial Model
Q. Consider a 3-year European call with K =$30 on a stock with
S0 =$30. Suppose that a time step is 1 year, and in each time step
the stock price moves either up or down by 10%. The risk-free
interest rate is 5%. What is the price of the call option?

14 / 30
N-Step Binomial Model
Q. Consider a 3-year European call with K =$30 on a stock with
S0 =$30. Suppose that a time step is 1 year, and in each time step
the stock price moves either up or down by 10%. The risk-free
interest rate is 5%. What is the price of the call option?
Answer: First, the option payoffs at each of 4 nodes are

f0 = max(30(1.1)0 (0.9)3 − 30, 0) = 0

f1 = max(30(1.1)1 (0.9)2 − 30, 0) = 0
f2 = max(30(1.1)2 (0.9)1 − 30, 0) = 2.67
f3 = max(30(1.1)3 (0.9)0 − 30, 0) = 9.93.

e 0.05×1 −0.9
The risk-neutral probability is p = 1.1−0.9
= 0.756. Then, the option price is
3  
X N
e −0.05×3 (0.756)j (1 − 0.756)N−j fj = 4.66
j
j=0

15 / 30
Pricing American Options

16 / 30
American Options

• In pricing American options, we should consider that the options can

be exercised early.

• In a similar way to pricing European options, we build a binomial tree

of stock price. Then, we start from final nodes and proceed backward.

• However, the option value at each node becomes the maximum of

1 the option value when delaying the exercise,

f = e −r ∆t [pfu + (1 − p)fd ]

2 the payoff when exercising now

17 / 30
American Options
e.g. Consider a 2-year American put with K =$52 on a stock with
S0 =$50. Suppose that a time step is 1 year, and in each time step
the stock price moves either up or down by 20%. The risk-free
interest rate is 5%. What is the price of the put option?

72
fuu
60 u
u fu d
50 48
f fud
d 40 u
fd d
32
fdd

18 / 30
American Options
• The risk-neutral probability is

e 0.05×1 − 0.8
p= = 0.6282.
1.2 − 0.8

• At final nodes, option payoffs are


 fuu = max(52 − 72, 0) = 0
f = max(52 − 48, 0) = 4
 ud
fdd = max(52 − 32, 0) = 20

19 / 30
American Options
• At top node in T = 1, the option price is
 
fu = max e −r ∆t [pfuu + (1 − p)fud ], 52 − 60}
| {z = $1.415.
| {z }
value if delay value if exercise

At bottom node in T = 1, the option price is

 
fd = max e −r ∆t [pfud + (1 − p)fdd ] , 52 − 40 = $12.

• At the initial node, the option price is

 
f = max e −r ∆t [pfu + (1 − p)fd ] , 52 − 50 = $5.090.

20 / 30
Determining u and d

21 / 30
Determining u and d

• We have studied how to price options when u and d are given in

binomial trees.

• If they are not given, how can we determine u and d?

• For this determination, we focus on the volatility of underlying asset.

• The volatility σ is the standard deviation of yearly returns on the stock.

• The basic idea is to choose u and d such that the volatility in the
binomial tree matches the volatility we see in data.

• First, how can we measure the volatility from data?

22 / 30
Determining u and d - Estimating Volatility

• Suppose that we have data of returns over ∆t horizon, not yearly

return.

• The standard deviation of r∆t can be related to the volatility as

follows: √
Std.Dev. (r∆t ) = σ ∆t.
where σ is the the standard deviation of r1-year .

23 / 30
Determining u and d - Estimating Volatility
• Suppose that we have the following historical prices of stock.
Date Price Return
0 P0
1 P1 r1 = ln(P1 /P0 )
2 P2 r2 = ln(P2 /P1 )
.. .. ..
. . .
T PT rT = ln(PT /PT −1 )

• Then, the standard deviation of (r1 , r2 , · · · , rT ) is the estimate for

daily return.

• Then the volatility σ can be found as follows:

r
1
Std.Dev (r1-day ) = σ .
365

24 / 30
Determining u and d

• Once the volatility σ is obtained from data, we want to construct a

binomial tree such that returns in the tree have the same volatility.
• This means that the return
√ over one step (= ∆t) should have the
standard deviation of σ ∆t.

• We can achieve this by choosing

√ √
u = eσ ∆t
and d = e −σ ∆t

in the binomial tree.

• Why does this choice of u and d work?

25 / 30
Determining u and d

• Suppose that the required return on the stock in the real world is α
α∆t
per annum. Then, the real probability p ∗ = e u−d−d .

• Using this real probability, we can compute the variance of return

Var(r ).

• We want to show that Var(r ) equals σ 2 ∆t under this particular

choice of u and d.
2
Var(r ) = E (r 2 ) − [E (r )]
2
= p ∗ u 2 + (1 − p ∗ )d 2 − e α∆t


= p ∗ (u 2 − d 2 ) + d 2 − e 2α∆t
= ...

26 / 30
Determining u and d

Var(r ) = . . .
e α∆t − d
= (u + d)(u − d) + d 2 − e 2α∆t
u−d
= e α∆t (u + d) − d(u + d) + d 2 − e 2α∆t
= (u + d)e α∆t − ud − e 2α∆t

√ √
• Now, let’s plug in u = e σ ∆t
and u = e −σ ∆t
.
√ √
Var(r ) = (e σ ∆t
+ e −σ ∆t
)e α∆t − 1 − e 2α∆t
= ...?

27 / 30
Determining u and d - Math Review

• Taylor series
• A function f (x) can be expressed as a sum of polynomials:

f ′′ (0) 2 f ′′′ (0) 3

f (x) = f (0) + f ′ (0)x + x + x + ...
2! 3!

• A Taylor series of e x is

1 2 1
ex = 1 + x + x + x 3 + ...
2! 3!

• When x is small, we can ignore higher-order terms.

28 / 30
Determining u and d

• Applying the Taylor series to the exponential terms, we can simplify

the variance.

• Here, we assume that ∆t is very small, so we ignore ∆t 3/2 , ∆t 2 , and

higher powers.
√
√ √ (σ ∆t )
2
• For instance, e σ ∆t
≈ 1 + σ ∆t + 2 .

• Then, the variance becomes

√ √
Var(r ) = (e σ + e −σ ∆t )e α∆t − 1 − e 2α∆t
∆t
√ √
≈ (2 + σ ∆t − σ ∆t + σ 2 ∆t)(1 + α∆t) − 1 − (1 + 2α∆t)
≈ σ 2 ∆t

29 / 30
Things To Do

• Read the textbook chapters 13.3 - 13.5, 13.7 - 13.8

• Assignment 8 (due on 11 April)

30 / 30