Chapter 3 Brownian Motion 3.2 Scaled random Walks 3.2.1 Symmetric Random Walk • To construct a symmetric random walk, we toss a fair coin (p, the probability of H on each toss, and q, the probability of T on each toss) 1 if j H, Xj = 1 if j = T, 3.2.1 Symmetric Random Walk • Define 𝑀0 = 0 • 𝑀𝑘 = 𝑘 𝑗=1 𝑋𝑗 k=1,2,….. , 3.2.2 Increments of the Symmetric Random Walk • A random walk has independent increments .If we choose nonnegative integers 0 = k k k , the random variables are independent 0 1 m • Each M ki1 M ki walk ki 1 j ki 1 X j is called increment of the random 3.2.2 Increments of the Symmetric Random Walk • Each increment M k M khas expected value 0 and variance ki 1 ki i 1 i M ki1 M ki ki1 ki1 X j X j j ki 1 j ki 1 ki 1 ki 1 1 1 = 1 ( 1) (0) 2 j ki 1 j ki 1 2 =0 3.2.2 Increments of the Symmetric Random Walk Var ( M ki1 M ki ) Var ( ki 1 X j ki 1 j ki 1 Var ( X j ki 1 (X j X i , i j) ) j ) ki 1 1 k j ki 1 i 1 ki 3.2.3 Martingale Property for the Symmetric Random Walk • Choose nonnegative integers k < l , then (M l F k ) = [( M l M k ) M k F k ] = [ M l M k F k ] +[M k F k ] = [M l M k F k ] + M k = [M l M k ] + M k = M k (( M l M k ) Fk ) 3.2.4 Quadratic Variation for the Symmetric Random Walk • The quadratic variation up to time k is defined to be M , M k = M j M j 1 k j 1 • Note : .this is computed path-by-path and .by taking all the one-step increments M j M j 1 along that path, squaring these increments, and then summing them 2 k 3.2.5 Scaled Symmetric Random Walk • To approximate a Brownian motion • Speed up time of a symmetric random walk • Scale down the step size of a symmetric random walk • Define the Scaled Symmetric Random Walk W (n) 1 (t ) = M nt n • If nt is not an integer, we define 𝑊 •𝑊 𝑛 𝑛 𝑡 by linear interpolation 𝑡 is a Brownian motion as n 3.2.5 Scaled Symmetric Random Walk • Consider • n=100 , t=4 3.2.5 Scaled Symmetric Random Walk • The scaled random walk has independent increments • If 0 = t0 t1 tm are such that each nt j is an integer, then (n) (n) (n) ( n) ( n) ( n) W ( t ) W ( t ) , W ( t ) W ( t ) , , W ( t ) W (tm1 ) 1 m 0 2 1 are independent • If 0 s t are such that ns and nt are integers, then W ( n ) (t ) - W ( n ) (s) 0, Var W ( n ) (t ) - W ( n ) (s) t s 3.2.5 Scaled Symmetric Random Walk • Scaled Symmetric Random Walk is Martingale W ( n ) (t ) F s W ( n ) (s ) • Let 0 s t be given and s , t are chosen so that ns and nt are integers W (t ) W ( s) F s W ( s) F s W (t ) W ( s) F s W ( s) n n n n W (t ) F s W (t ) W (s) W (s) F s n n n n n n n n n n W (t ) W ( s) W ( s) W ( s ) 3.2.5 Scaled Symmetric Random Walk • Quadratic Variation In general, for t 0 such that nt is an integer W (n) j (n) j 1 t W W n n j 1 nt (n) ,W (n) 2 nt 1 1 = Xj t j 1 n j 1 n nt 2 3.2.6 Limiting Distribution of the Scaled Random Walk • We fix the time t and consider the set of all possible paths evaluated at that time t • Example 1 • Set t = 0.25 and consider the set of possible values of W (100) 0.25 M 25 10 • We have values: -2.5,-2.3,…,-0.3,-0.1,0.1,0.3,…2.3,2.5 • The probability of this is 25 25! 1 W (100) 0.25 0.1 0.1555 13!12! 2 3.2.6 Limiting Distribution of the Scaled Random Walk • The limiting distribution of W (100) 0.25 W (100) 0.25 0 Var W (100) 0.25 0.25 • Converges to Normal 3.2.6 Limiting Distribution of the Scaled Random Walk • Given a continuous bounded function g(x) g W (100) 0.25 2 2 g ( x )e 2 x 2 dx 3.2.6 Limiting Distribution of the Scaled Random Walk • Theorem 3.2.1 (Central limit) Fix t 0. As n , the distribution of the scaled random walk W ( n ) (t ) evaluated at time t converges to the normal distribution with mean 0 and variance t 藉由MGF的唯一性來判斷r.v.屬於何種分配 3.2.6 Limiting Distribution of the Scaled Random Walk • Let f(x) be Normal density function with mean=0, variance=t 1 f ( x) e 2 t x2 2t In general, x (u ) e u 2u 2 Now, we have x (u ) e 2 u 2t 2 2 3.2.6 Limiting Distribution of the Scaled Random Walk • If t is such that nt is an integer, then the m.g.f. for W n (t ) is n u e n uW ( t ) u exp M nt n u nt nt u exp X j exp Xj n j 1 n j 1 u exp Xj n j 1 nt 1 e j 1 2 nt u n 1 e 2 u n X i X j i j 1 e 2 u n 1 e 2 u n nt 3.2.6 Limiting Distribution of the Scaled Random Walk • To show that 1 lim n (u) lim( e n n 2 u n 1 e 2 u n ) nt (u) e 1 2 u t 2 • Then, 1 lim lnn (u) lim nt ln( e n n 2 u n 1 e 2 u n 1 2 ) u t 2 3.2.6 Limiting Distribution of the Scaled Random Walk 1 key : Let x n 1 ux 1 ux ln e e 0 2 2 lim ln n u lim t ( 型 L'Hopital's rule) 2 n x 0 x 0 u ux u ux e e ux ux tu e e 0 2 2 lim t lim ux ux ( 型) x 0 x 0 x (e e 1 ux 1 ux 2 ) 0 2 x( e e ) 2 2 tu ueux ueux tu u u 1 2 lim ux ux ut ux ux 2 x0 e e x(ue ue ) 2 1 1 2 3.2.7 Log-Normal Distribution as the Limit of the Binomial Model • The Central Limit Theorem, (Theorem3.2.1), can be used to show that the limit of a properly scaled binomial asset-pricing model leads to a stock price with a log-normal distribution • Assume that n and t are chosen so that nt is an integer • Up factor to be un 1 • Down factor to be n dn 1 n • is a positive constant 3.2.7 Log-Normal Distribution as the Limit of the Binomial Model • The risk-neutral probability and we assume r=0 1 r dn 1 n p 2 un d n 2 n ~ un 1 r 1 n q 2 un d n 2 n ~ 3.2.7 Log-Normal Distribution as the Limit of the Binomial Model • The stock price at time t is determined by the initial stock price S(0) and the result of first nt coin tosses • H nt: the sum of the number of heads • Tnt : the sum of the number of tails nt H nt Tnt 3.2.7 Log-Normal Distribution as the Limit of the Binomial Model • The random walk M nt is the number of heads minus the number of tails in these nt coin tosses M nt H nt Tnt nt H nt Tnt 1 H nt = 2 nt M nt T 1 nt M nt nt 2 3.2.7 Log-Normal Distribution as the Limit of the Binomial Model • We wish to identify the distribution of this random variables as S n t S (0)u d H nt n n Tnt n S (0) 1 n 1 nt M nt 2 1 n 1 nt M nt 2 1 2 S (t ) S (0) exp{W (t ) t} as n 2 • Where W(t) is a normal random variable with mean 0 amd variance t 3.2.7 Log-Normal Distribution as the Limit of the Binomial Model • We take log for equation 1 1 log S n (t ) log S (0) (nt M nt ) log(1 ) (nt M nt ) log(1 ) 2 n 2 n • To show that it converges to distribution of 1 2 log S (t ) log S (0) W (t ) t 2 3.2.7 Log-Normal Distribution as the Limit of the Binomial Model • Taylor series expansion f ( x) f n 0 (n) (a) n ( x a) n! • Expansion at 0 1 2 3 f ( x) f (0) f ' (0) x f " (0) x O( x ) 2 • Let log(1+x)=f(x) 1 2 3 log(1 x) x x O( x ) 2 3.2.7 Log-Normal Distribution as the Limit of the Binomial Model 1 log S (t ) log S (0) (nt M nt )( O(n )) 2 n 2n 2 3 2 1 (nt M nt )( O (n )) 2 n 2n 2 log S (0) nt( Note : W (n) 1 (t ) M nt n 2 2n 3 2 O(n )) M nt ( 3 2 n 1 2 (n) log S (0) t ntO(n ) W (t ) 2 ) 3 2 3.2.7 Log-Normal Distribution as the Limit of the Binomial Model • Then 3 2 1 2 ( n) lim log S n (t ) lim log S (0) t W (t ) O(n ) n n 2 1 2 log S (0) t W (t ) 2 • Hence 1 2 S (t ) S (0) exp{W (t ) t} 2