Anda di halaman 1dari 7

Math 621, Mathematics Finance I

Lecture I. The Binomial No-Arbitrage Pricing Model


Ruoting Gong

1 One Period Binomial Model


1.1 The Model Settings
A stock with time-zero price S0 > 0.

A coin is tossed, with the probability p of a head, 0 < p < 1. Denote by q = 1p, the probability
of a tail.

The outcome of the coin toss determines the price of the stock price at time one

S1 (H), with probability p
S1 =
S1 (T ), with probability q = 1 p

Set u = S1S(H)
0
(up factor), d = S1 (T )
S0 (down factor). Without loss of generality, assume d < u.
Usually we set d = u1 .

A money market with interest rate (for both borrowing and investing) r 0. Mathematically
we only requires r 1.

1.2 No-Arbitrage Condition


Definition 1.1 (Arbitrage) A trading strategy is called an arbitrage if it begins with no money, has
zero probability of losing money, and has a positive probability of making money.

Theorem 1.1 In the one-period binomial model,

No Arbitrage 0 < d < 1 + r < u. (1.1)

Proof: : Suppose that 1 + r d < u. We can build an arbitrage strategy as follows: begin
with zero wealth, borrow S0 from the money market and buy one share of stock at time zero. Since
1 + r d < u, at time one,

S1 (H) = uS0 > (1 + r)S0 , S1 (T ) = dS0 (1 + r)S0 .

The stock at time one is worth enough to pay off the money market debt and has a positive probability
of being worth strictly more.

1
Next, if d < u 1 + r, there exists again an arbitrage strategy: sell the stock short and invest S0
in the money market. Since d < u 1 + r, at time one,

S1 (H) = uS0 (1 + r)S0 , S1 (T ) = dS0 < (1 + r)S0 .

The cost of replacing the stock at time one will be less than or equal to the value of the money market
investment.
: We need to show if the initial wealth X0 = 0, for any choice of 0 (the number of stocks held
at time zero), at time one the wealth

X1 = 0 S1 + (1 + r)(0 S0 )

cannot be strictly positive with positive probability unless it is strictly negative with positive proba-
bility as well. The detail proof will be left as an exercise. 2

1.3 No-Arbitrage Pricing of Derivative Securities


We make the following assumptions:

Shares of stock can be subdivided for sales or purchase. This is essentially satisfied in practice
because option pricing and hedging typically involves lots of options.

The interest rates for investing and borrowing are the same. This is close to being true for large
institutions.

Zero bid-ask spread: the purchase price and the selling price of the stock are the same. This
is not satisfied in practice, but can be ignored sometimes when not too much trading is taking
place.

At any time, the stock can take only two possible values in the next period. We shall develop
more sophisticated models that are not tied to this assumption.

In general, we consider a derivative security under the one-period binomial model. That is, a
security paying V1 (H) (respectively V1 (T )) at time one, if the coin toss results in head (respectively
tail). The next theorem gives the no-arbitrage price of a derivative security at time zero.

Theorem 1.2 Assume that the no-arbitrage condition (1.1) is valid. The price V0 at time zero for a
derivative security is given by
1
V0 = [pV1 (H) + qV1 (T )], (1.2)
1+r
where
1+rd u1r
p = , q = 1 p = . (1.3)
ud ud
Proof: To determine the time-zero price V0 , we will find the replicating portfolio. At time zero,
suppose we begin with initial wealth X0 = V0 , and buy 0 shares of stock, leaving with a cash
position V0 0 S0 in the money market. At time one, the value of our portfolio of stock and money
market account is

X1 = 0 S1 + (1 + r)(V0 0 S0 ) = (1 + r)V0 + 0 (S1 (1 + r)S0 ).

2
We want to find V0 and 0 such that X1 (H) = V1 (H), X1 (T ) = V1 (T ), that is,
 
1 1
V0 + 0 S1 (H) S0 = V1 (H), (1.4)
1+r 1+r
 
1 1
V0 + 0 S1 (T ) S0 = V1 (T ). (1.5)
1+r 1+r

We first choose 0 < p < 1, such that


1
S0 = [pS1 (H) + (1 p)S1 (T )], (1.6)
1+r
Since S1 (H) = uS0 and S1 (T ) = dS0 , (1.3) follows directly from (1.6). By adding p(1.4) to (1
p)(1.5), we obtain
 
1 1
V0 + 0 [pS1 (H) + (1 p)S1 (T )] S0 = [pV1 (H) + (1 p)V1 (T )],
1+r 1+r

and (1.2) follows from (1.6). Moreover, by subtracting (1.5) from (1.4),

V1 (H) V1 (T )
0 = . (1.7)
S1 (H) S1 (T )

Therefore, at time zero, if the agent begins with wealth X0 = V0 given in (1.2) and buys 0 shares
of stock, where 0 is given by (1.7), then he has hedged a short position in the derivative security.
Hence, the derivative security that pays V1 at time one should be priced as in (1.2) at time zero. 2

Remark 1.1 (a) To hedge a long position in the derivative security, the number of shares of the
underlying stock held by the agent is the negative of (1.7).

(b) The numbers p and q given by (1.3) are both in (0, 1) because of the no-arbitrage condition (1.1).

(c) p and q are so-called risk-neutral probabilities. Under the risk-neutral probabilities, the average
rate of growth of the stock is equal to the rate of growth of an investment in the money market
(see (1.6)). Hence, they make the mean rate of growth of any portfolio of stock and money
account appear to equal the rate of growth of the money market.

(d) The risk-neutral probabilities p and q are not the actual probabilities p and q. Under the actual
probabilities, the average rate of growth of stock is typically strictly greater than the rate of growth
of an investment in the money market. Otherwise, no one would want to incur the risk associated
with investing in the stock. Thus, p and q should satisfy
1
S0 < [pS1 (H) + qS1 (T )].
1+r

(e) The equation (1.2) is called the risk-neutral pricing formula of the derivative security for the
one-period binomial model.

3
2 Multi-period Binomial Model
In this section, we will use the same notations as before:

S0 : Stock price at time zero.

u and d: up factor and down factor.

r: interest rate of the money market.

Assume the no-arbitrage condition (1.1), a three period binomial model can be defined as follows:

Time-zero price: S0 .

Time-one price depends on the first coin toss: S1 (H) = uS0 , S1 (T ) = dS0 .

Time-two price depends on the first two coin tosses:

S2 (HH) = u2 S0 , S2 (HT ) = S2 (T H) = udS0 , S2 (T T ) = d2 S0 .

Time-three price depends on the first three coin tosses:

S3 (HHH) = u3 S0 , S3 (HHT ) = S3 (HT H) = S3 (T HH) = u2 dS0 ,


S3 (T T T ) = d3 S0 , S3 (HT T ) = S3 (T HT ) = S3 (T T H) = ud2 S0 , .

Similarly, we can define the general N -step binomial model, for any integer N N.
To derive the no-arbitrage prices of derivative securities under a general N -step binomial model,
we will need the notations of the following stochastic processes:

Stock prices: S0 , S1 , S2 , . . . , SN .

Values of derivative security: V0 , V1 , V2 , . . . , VN .

Values of hedging portfolio: X0 , X1 , X2 , . . . , XN .

Number of shares of stocks held in the hedging portfolio: 0 , 1 , 2 , . . . , N .

By stochastic process, we mean a sequence of random variables indexed by time period. These quan-
tities are random since they depend on the random outcomes of coin tosses (S0 , V0 , X0 and 0 are
deterministic numbers). We will also denote the outcomes of coin tosses by w1 , w2 , . . . , wN , where
wi = H or T , i = 1, . . . , N . Notice that the values of the hedging portfolio is determined by

Xn+1 = n Sn+1 + (1 + r)(Xn n Sn ), n = 0, . . . , N 1. (2.1)

Theorem 2.1 Consider an N -period binomial model satisfying the no-arbitrage condition (1.1). Let
p and q be the risk-neutral probabilities given in (1.3). Let VN be the payoff of a derivative security
at time N , depending on the first N coin tosses w1 , w2 , . . ., wN . Define recursively backward in time
the sequence of random variables VN 1 , VN 2 , . . . , V0 by
1
Vn (w1 . . . wn ) = [pVn+1 (w1 . . . wn H) + qVn+1 (w1 . . . wn T )], n = 0, . . . , N 1. (2.2)
1+r

4
Next, define

Vn+1 (w1 . . . wn H) Vn+1 (w1 . . . wn T )


n (w1 . . . wn ) = , n = 0, . . . , N 1. (2.3)
Sn+1 (w1 . . . wn H) Sn+1 (w1 . . . wn T )

If we set X0 = V0 and define recursively forward in time the portfolio values X1 , X2 , . . ., XN by (2.1),
then we will have

Xn (w1 . . . wn ) = Vn (w1 . . . wn ), for all w1 . . . wn , n = 1, . . . N. (2.4)

Definition 2.1 For n = 1, . . . , N , the random variable Vn (w1 . . . wn ) in Theorem 2.1 is the no-
arbitrage price of the derivative security at time n, if the outcomes of the first n tosses are w1 . . . wn .
The no-arbitrage price of the derivative security at time zero is V0 .

Proof of Theorem 2.1: We will prove (2.4) by induction. The case of n = 0 is given by definition
X0 = V0 . Suppose that (2.4) is valid for some n < N and all possible outcomes w1 , . . . , wn , we need
to derive (2.4) for n + 1 and any arbitrary outcome w1 , . . . , wn+1 .
Fix arbitrary w1 , . . . , wn , using (2.1)-(2.3) and the induction hypothesis,

Xn+1 (w1 . . . wn H) = n (w1 . . . wn )Sn+1 (w1 . . . wn H)+(1+r)(Xn (w1 . . . wn )n (w1 . . . wn )Sn (w1 . . . wn ))
= n (w1 . . . wn )Sn (w1 . . . wn )(u 1 r) + (1 + r)Xn (w1 . . . wn )
Vn+1 (w1 . . . wn H)Vn+1 (w1 . . . wn T )
= Sn (w1 . . . wn )(u1r)+(1+r)Xn (w1 . . . wn )
Sn+1 (w1 . . . wn H)Sn+1 (w1 . . . wn T )
Vn+1 (w1 . . . wn H)Vn+1 (w1 . . . wn T )
= Sn (w1 . . . wn )(u1r)+(1+r)Xn (w1 . . . wn )
(u d)Sn (w1 . . . wn )
= q (Vn+1 (w1 . . . wn H)Vn+1 (w1 . . . wn T ))+ pVn+1 (w1 . . . wn H)+ qVn+1 (w1 . . . wn T )
= Vn+1 (w1 . . . wn H).

Similarly, we can show that

Xn+1 (w1 . . . wn T ) = Vn+1 (w1 . . . wn T ),

which completes the proof. 2

Examaple 2.1 (Path-dependent Options) Suppose that S0 = 4, u = 2, d = 0.5, r = 0.25. Then


p = q = 0.5. Consider a lookback option that pays off

V3 = max Sn S3
0n3

at time three. Then

V3 (HHH) = S3 (HHH)S3 (HHH) = 3232 = 0, V3 (HHT ) = S2 (HH)S3 (HHT ) = 168 = 8,


V3 (HT H) = S1 (H)S3 (HT H) = 88 = 0, V3 (T HH) = S3 (T HH)S3 (T HH) = 88 = 0,
V3 (HT T ) = S1 (H)S3 (HT T ) = 82 = 6, V3 (T HT ) = S2 (T H)S3 (T HT ) = 42 = 2,
V3 (T T H) = S0 S3 (T T H) = 42 = 2, V3 (T T T ) = S0 S3 (T T T ) = 40.5 = 3.5.

5
Using the backward recursion (2.2),
   
4 1 1 4 1 1
V2 (HH) = V3 (HHH) + V3 (HHT ) = 3.2, V2 (HT ) = V3 (HT H) + V3 (HT T ) = 2.4,
5 2 2 5 2 2
   
4 1 1 4 1 1
V2 (T H) = V3 (T HH) + V3 (T HT ) = 0.8, V2 (T T ) = V3 (T T H) + V3 (T T T ) = 2.2,
5 2 2 5 2 2

and then
   
4 1 1 4 1 1
V1 (H) = V2 (HH) + V2 (HT ) = 2.24, V1 (T ) = V2 (T H) + V2 (T T ) = 1.2,
5 2 2 5 2 2

and finally  
4 1 1
V0 = V1 (H) + V1 (T ) = 1.376.
5 2 2
To hedge the short position of this lookback option, one need to set up a portfolio by buying

V1 (H) V1 (T ) 2.24 1.2 13


0 = = = 0.1733
S1 (H) S1 (T ) 82 75

shares of stock and investing 1.376 0.1733 4 = 0.6827 in the money market at 25% interest.

3 Computational Considerations
The amount of computation required by a naive implementation of the derivative security pricing
algorithm given in Theorem 2.1 grows exponentially with the number of periods. The binomial models
used in practice often have more than 100 periods, and there are 2100 possible outcomes for a sequence
of 100 coin tosses. An algorithm that begins with tabulating 2100 values for V100 is not computationally
practical.
The following two examples illustrate a computationally efficient manner to organize the algorithm
given in Theorem 2.1.

Examaple 3.1 In the model given in Example 2.1, consider the problem of pricing a European put
with strike price K = 5, expiring at time N . The payoff of the option is given by VN = (5 SN )+ .
When N = 3, it can be tabulated as:

V3 (HHH) = 0, V3 (HHT ) = V3 (HT H) = V3 (T HH) = 0,


V3 (HT T ) = V3 (T HT ) = V3 (T T H) = 3, V3 (T T T ) = 4.5.

Observe that the payoffs have the same value when the outcomes of tosses have the same combination.
Indeed, the stock price SN , which determines the payoff of the European put, remains the same if the
combinations of the N coin tosses are the same. Although there are 23 = 8 possible permutations of
the coin tosses, there are only 3 + 1 = 4 possible combinations of the outcomes, and thus 4 possible
values of the stock price at time three. Therefore, if we let v3 (s) be the payoff of the option at time
three when the stock price is S3 = s, then we can tabulate the values of v3 as:

v3 (32) = 0, v3 (8) = 0, v3 (2) = 3, v3 (0.5) = 4.5.

6
Next, we consider the the price V2 of the European put at time 2. From Theorem 2.1,
2
V2 (1 2 ) = [V3 (1 2 H) + V3 (1 2 T )] . (3.1)
5
According to the discussion above, the value of the right hand side of (3.1) varies if and only if the
combination of 1 2 varies. Therefore, the value of V2 is determined by different combinations of
1 2 , or equivalently, different values of S2 .
In general, let vn (s) be the payoff of the option at time n when the stock price is Sn = s (Essen-
tially, it is due to the fact that {Sn }N n=0 is a Markov Process). The argument of vn would
only range over the n + 1 possible stock prices at time n, while the argument of Vn would range over
2n possible outcomes of the coin tosses. The algorithm in Theorem 2.1 can then be simplified as:
  
2 1
vn (s) = vn+1 (2s) + vn+1 s , n = 0, . . . , N 1,
5 2

and the number of shares of the stock that should be held by the replicating portfolio is

vn+1 (2s) vn+1 12 s



n (s) = , n = 0, . . . , N 1.
2s 21 s

Examaple 3.2 Consider the lookback option of Example 2.1. At each time n, the price of the option
is a function of the stock price Sn and the maximum stock price Mn = max0kn Sk to date (since
{Sn , Mn }N
n=0 is a bivariate Markov Process). The algorithm of Theorem 2.1 can be rewritten in
terms of the functions vn as
  
2 1
vn (s, m) = vn+1 (2s, m (2s)) + vn+1 s, m ,
5 2

and the number of shares of stock that should be held by the replication portfolio is

vn+1 (2s, m (2s)) vn+1 ( 12 s, m)


n (s, m) = .
2s 12 s

Anda mungkin juga menyukai