STK4510 - Mandatory Assignment II Sindre Froyn Exercise 3 We are working with two stocks with the dynamics: n o p dS1 (t) = µ1 S1 (t)dt + S1 (t) ρσ1 dB1 (t) + 1 − ρ2 σ1 dB2 (t) dS2 (t) = µ2 S2 (t)dt + S2 (t)σ2 dB1 (t) and the price of a spread option paying S2 (T ) − S1 (T ) (a) + at time T . We begin by finding expressions for S1 (t) and S2 (t). For S1p (t) we simplify the notation by defining: S1 (t) = St , µ1 = µ, σ1 = σ and γ = 1 − ρ2 . We have the dynamics dSt = µSt dt + ρσSt dB1 + γσSt dB2 . (1) To solve this we make a guess which will provide some insight to the solution: St = S0 exp {µt + ρσB1 + γσB2 } . We find the dynamics for this process by using the multi dimensional Ito’s formula: f (t, x1 , x2 ) = S0 exp {µt + ρσx1 + γσx2 } where n n X ∂f ∂f 1 X ∂2f df = dBi + dt. dt + ∂t ∂xi 2 i=1 ∂x2i i=1 In the last sum we don’t need to consider any cross double derivatives, and we get dt because we are using Brownian motion, which means we have the relation: dt, i = j dBi dBj = 0, i 6= j 1 1 df (t, B1, B2 ) = µSt dt + ρσSt dB1 + γσSt dB2 + ρ2 σ 2 St dt + γ 2 σ 2 St dt 2 2 h i 1 1 dSt = St µ + ρ2 σ 2 + γ 2 σ 2 dt + ρσSt dB1 + γσSt dB2 2 2 1 Comparing with the dynamics we are supposed to get in (1), we see that we have two excessive terms in the dt-term. These would be eliminated if we included −1/2(ρ2 σ 2 + γ 2 σ 2 )t in our guess. We also note: − σ2 2 σ2 σ2 (ρ + γ 2 ) = − (ρ2 + 1 − ρ2 ) = − . 2 2 2 Hence, the correct solution for the SDE is, reintroducing the original notation: p σ12 2 t + ρσ1 B1 (t) + 1 − ρ σ1 B2 (t) . S1 (t) = S1 (0) exp µ1 − 2 For S2 (t) we have a normal Geometric Brownian motion, and by Ito’s formula we can easily verify the solution: σ22 S2 (t) = S2 (0) exp µ2 − t + σ2 B1 (t) . 2 Writing the price P at time 0 as an expectation of the price, is simply: + −rT P = e E S2 (T ) − S1 (T ) We are going to write this under the risk neutral measure Q, and since we have two stocks and two Brownian motions, we use Multi dimensional Girsanov’s Theorem. B1 W1 −1 (µ1 − r)t +Σ = (µ2 − r)t B2 W2 p Recalling that γ = 1 − ρ2 : 1 0 ρσ1 γσ1 −1 σ2 =⇒ Σ = 1 Σ= − γσρ 2 σ2 0 γσ1 Multiplying the matrix Σ−1 in the expression, and shifting terms gives us the following expressions: µ2 − r dt σ2 µs − r µ1 − r dt + ρ dt dB2 = dW2 − γσ1 γσ2 dB1 = dW1 − Now we return to the dynamics for S1 and S2 , and use the new Q-Brownian motions instead, beginning with S2 : µ2 − r dS2 = µ2 S2 dt + S2 σ2 dW1 − dt = rS2 dt + σ2 S2 dW1 σ2 2 µ2 − r µ1 − r µ2 − r dS1 = µ1 S1 dt + S1 ρσ1 dW1 − dt + γσ1 dW2 − dt + ρ dt σ2 γσ1 γσ2 µ − r µ − r 2 1 + r)dt + ρσ dt dt + γσ1 dW2 − (µ = µ1 S 1 1 1 dt + S1 ρσ1 dW1 − ρσ1 σ2 σ2 = rS1 dt + ρσ1 S1 dW1 + γσ1 S1 dW2 The dynamics under the Q measure have the same structure as under the physical measure. We can solve the SDE’s under Q in the same manner, by just replacing the appropriate variables. We write the solutions with T for convenience. σ12 S1 (T ) = S1 (0) exp rT − T + ρσ1 W1 (T ) + γσ1 W2 (T ) 2 σ22 S2 (T ) = S2 (0) exp rT − T + σ2 W1 (T ) 2 Now we can write the price of the spread option under Q. σ22 −rT P = e EQ S2 (0) exp rT − T + σ2 W1 (T ) − 2 + σ12 S1 (0) exp rT − T + ρσ1 W1 (T ) + γσ1 W2 (T ) 2 We factor out erT which cancels out the discounting factor, and S2 (0) from both terms. 2 σ2 P = S2 (0)EQ exp − T + σ2 W1 (T ) − 2 2 + S1 (0) σ1 exp − T + ρσ1 W1 (T ) + γσ1 W2 (T ) S2 (0) 2 Finally we factor out the remaining exponential in the first term. This must remain within the expectation as it contains a Brownian component. σ2 σ2 S1 (0) exp (− 1 + 2 )T + (ρσ1 − σ2 )W1 (T ) (2) P = S2 (0)EQ 1 − S2 (0) 2 2 + 2 p σ2 2 + 1 + ρ σ1 W2 (T ) (3) × exp − T + σ2 W1 (T ) 2 We can factor out of the max-function, because if the difference S2 (T )−S1 (T ) is positive, we have only reformulated the expression, and if the difference is 0 or negative, the max-function becomes 0 and so does the entire expression. 3 (b) e Brownian motion. Using Girsanov’s theorem, we introduce a Q f1 (t) = −σ2 dt + dW1 (t) ⇐⇒ dW1 (t) = σ2 dt + dW f1 (t) dW Writing in integral form up to T , we can calculate the deterministic part. Z T f1 (T ) =⇒ W1 (T ) = σ2 T + W f1 (T ) W1 (T ) = σ2 dt + W 0 We perform som intermediate calculations. For the expression in the exponential in (2): f1 (T ) =⇒ (ρσ1 − σ2 )W1 (T ) = (ρσ1 σ2 − σ22 )T + (ρσ1 − σ2 )W σ12 σ22 σ2 σ2 f1 (T ) + )T + (ρσ1 − σ2 )W1 (T ) = (− 1 + ρσ1 σ2 − 2 )T + (ρσ1 − σ2 )W 2 2 2 2 For the exponent in (3), we observe that at time T , W1 (T ) ∼ N (0, T ), or 2 µW = 0 and σW = T , and by independence 1 2 1 2 2 σ T . E[exp(σ2 W1 (T ))] = exp µW σ2 + σW σ2 = exp 2 2 2 (− This cancels against the first term in the exponent from (3), so we get exp(0) = 1 and this factor vanishes, and we are left with: S1 (0) 1 f1 (T ) P = S2 (0)EQe 1 − exp − (σ12 − 2ρσ1 σ2 + σ22 )T + (ρσ1 − σ2 )W S2 (0) 2 + p + 1 + ρ2 σ1 W2 (T ) (c) We define: a = (ρσ1 − σ2 ) and b = σ1 f1 (T ) + bW2 (T ) with: distribution aW p h i f E aW1 (T ) + bW2 (T ) = 0, 1 − ρ2 . We have the centered normal h i2 f E aW1 (T ) + bW2 (T ) = 0. We calculate the variance: i h i2 h i h f1 (T ) + bW2 (T ) . f1 (T ) + bW2 (T ) = E (aW f1 (T ) + bW2 (T ))2 − E aW Var aW 4 The last term is simply zero, and we get: h i h i h i h i f1 (T )+bW2 (T ) = a2 E W f12 (T ) +ab E W f1 (T )W2 (T ) +b2 E W22 (T ) . Var aW | {z } | {z } | {z } =0 =T =T 2 2 2 2 2 2 2 ρ2 σ ρ2 σ = (a2 +b2 )T = 1 +σ2 −2ρσ1 σ2 +σ1 − 1 T = (σ1 +σ2 −2ρσ1 σ2 )T := σ T. To calculate the price P we observe that we have the form of a put option with strike price K = 1 and stock price ST := S1 (T )/S2(T ). By Black-Scholes formula, the price of a standard put option E[(K − ST )+ ] at time t = 0 is given by, using put/call parity: P0p = S0 − Ke−rT − P0c = S0 − Ke−rT − S0 Φ(u1 ) + Ke−rT Φ(u2 ), where P0c is the call option price. Replacing with our parameters and multiplying with S2 (0) and u1 and u2 as given as in theorem 4.3 with σ as defined above, we get: S1 (0) S1 (0) −rT −rT P = S2 (0) −e − Φ(u1 ) + e Φ(u2 ) S2 (0) S2 (0) S1 (0) −rT 1 − Φ(u1 ) − e 1 − Φ(u2 ) = S2 (0) S2 (0) −rT 1 − Φ(u2 ) = S1 (0) 1 − Φ(u1 ) − S2 (0)e = S1 (0)Φ(−u1 ) − S2 (0)e−rT Φ(−u2 ). 5