FEC FINANCIAL ENGINEERING CLUB PRICING EUROPEAN OPTIONS AGENDA ο Stochastic Processes ο Stochastic Calculus ο Black-Scholes Equation STOCHASTIC PROCESSES A SIMPLE PROCESS ο Let ππ‘ = 1 with probability π = 0.5 and ππ‘ = −1 with probability 1 − π = .5 (for all t) and consider the symmetric random walk, ππ‘ = π‘π=0 ππ ο Assume that ππ ’s are i.i.d. ο π0 = 0 ο Both π = (ππ‘ : π‘ ∈ β ∪ 0 ) and π = (ππ‘ : π‘ ∈ β ∪ 0 ) are random processes ο A random/stochastic process is (vaguely) just a collection of random variables ο They could be i.i.d. ο They may be correlated—they may even have different distributions ο There is no general theory/application for random processes until more context and structure is applied A SIMPLE PROCESS ο Note that ππ‘ ’s are iid with πΈ ππ‘ = .5 ∗ 1 + .5 ∗ −1 = 0 , ∀π‘ and Var ππ‘ = πΈ ππ‘ 2 = .5 ∗ 1 ο Then πΈ ππ‘ = πΈ π‘ π=0 ππ Var ππ‘ = Var = 2 + .5 ∗ −1 π‘ π=0 πΈ[ππ ] π‘ π=0 ππ = 2 = 1, ∀π‘ = 0 and π‘ π=0 πππ(ππ ) =π‘ A SIMPLE PROCESS ο Generally, we care about the increments of a process: π‘ ππ‘ − ππ = ππ π=π +1 So that πΈ ππ‘ − ππ = πΈ π‘ π=π +1 ππ = 0 , and πππ ππ‘ − ππ = πππ π‘ π=π +1 ππ =π‘−π ο The symmetric random walk is defined to have independent increments ο A process X is said to have independent increments if, for 0 = π‘0 < π‘1 < … < π‘π the increments ππ‘1 − ππ‘0 , ππ‘2 − ππ‘1 , … , ππ‘π − ππ‘π−1 are independent QUADRATIC VARIATION ο Define the quadratic variation of a sequence π up to time π‘ as [π, π]π‘ = π‘π=1(ππ − ππ−1 )2 ο This is a path-dependent measure of variation (thus it is random) ο For some unique processes, it may not be random ο For our symmetric random walk, note that a one step increment, ππ − ππ−1 , is either 1 or −1. Thus [π, π]π‘ = π‘ (ππ − ππ−1 π=1 )2 = π‘ π=1 1=π‘ SCALED SYMMETRIC RANDOM WALK Let π π π‘ = πππ‘ π be a scaled symmetric random walk π ο If π ∗ π‘ is not an integer, π integers of π ∗ π‘ π‘ is interpolated between the two neighboring ο Like a the symmetric r.w., the scaled symmetric r.w. has independent increments ο πΈπ ο π π‘ −π [π (π) , π (π) ]π‘ = π π ππ‘ π=1 =0 π π πππ π π π −π π π−1 π π 2 π‘ −π = ππ‘ π=1 π π ππ 2 π =π‘−π = ππ‘ 1 π=1 π =π‘ BROWNIAN MOTION ο By the central limit theorem π Brownian motion π π‘ π.π . π π‘ as π → +∞, where π π‘ is a Properties of B.M. 1) π π 0 = 0 = 1 2) π has independent increments 3) π π‘ − π π ~π(0, π 2 π‘ − π ) for π‘ ≥ π (we have been using B.M. with π = 1) 4) π π π‘ ππ ππππ‘πππ’ππ’π = 1 BROWNIAN MOTION Ex) What is πΈ π(π‘) π(π )] assuming π ≤ π‘ (suppose W has parameter σ) πΈπ π‘ π π ] = πΈ (π(π‘) − π π ) + π π π π ] = πΈ π π‘ − π π ) |π π ] + πΈ[π(π ) π π ] = 0 + π(π ) = π(π ) Ex) What is π π π , π‘ = πΈ π π π(π‘) ? π π π , π‘ = πΈ π π π(π‘) = πΈ π π − π 0 = πΈ π π −π 0 = π π 2 + 0 ∗ 0 = π π 2 2 + πΈ π π −π 0 ππ‘ − ππ , ππ independent → π is a martingale π π‘ −π π +π π −π 0 π π‘ −π π BROWNIAN MOTION ο Note that B.M. is a function and not a sequence of random variables and so our definition of quadratic variation must be altered: Let π be a partition of the interval 0, π : π‘0 , π‘1 , … , π‘π with 0 = π‘0 < π‘1 < … < π‘π = π. Let π = max π‘π+1 − π‘π . For a π function π π‘ , the quadratic variation of π up to time T is π, π π = lim π →0 π−1 π=0 [π π‘π+1 − π(π‘π )]2 BROWNIAN MOTION AND QUADRATIC VARIATION ο Note if π has a continuous derivative, π−1 π=0 [π π‘π+1 − π(π‘π )]2 Then π, π π ≤ lim = π−1 π=0 ≤ π π →0 π π′ π−1 π=0 π−1 π=0 = lim π ∗ lim π →0 =0 ∗ π →0 π ′ |π 0 2 π₯π∗ π‘π+1 − π‘π 2 π ′ π₯π∗ π′ π₯π∗ π−1 π=0 2 π‘ | ππ‘ = 0 2 2 π′ (by MVT) π‘π+1 − π‘π π‘π+1 − π‘π π₯π∗ 2 π‘π+1 − π‘π BROWNIAN MOTION AND QUADRATIC VARIATION ο For a B.M. π(π‘), consider the random variable π π‘π+1 − π π‘π πΈ πππ π π‘π+1 − π π‘π π π‘π+1 − π π‘π − 2(π‘π+1 − π‘π ) πΈ 2 2 = π‘π+1 − π‘π 2 =πΈ π π‘π+1 − π π‘π π π‘π+1 − π π‘π 2 2 −(π‘ π+1 − π‘π ) 2 = πΈ π π‘π+1 − π π‘π 4 + (π‘π+1 − π‘π )2 = 3(π‘π+1 − π‘π )2 − 2(π‘π+1 − π‘π )2 + (π‘π+1 − π‘π )2 = π(ππ+π − ππ )π BROWNIAN MOTION AND QUADRATIC VARIATION π π‘π+1 −π π‘π . Choose π large so that π‘π π‘π+1 − π‘π 2 ππ+1 2 π‘π+1 − π(π‘π )] = π π ο Let ππ+1 = [π ο Then π, π π = lim lim π→∞ 2 π−1 ππ+1 π=0 π π →0 π−1 π=0 [π π‘π+1 − π(π‘π = )]2 ππ . Then π‘π+1 π = lim π ∗ π→∞ = 1 by LLN. ο Conclusion π, π π = lim π →0 ο π πΎ π π πΎ = π π ο Similarly, ππ‘ππ‘ = 0 and ππππ‘ = 0 π−1 π=0 [π π‘π+1 − π(π‘π )]2 = π − π‘π = 2 π−1 ππ+1 π=0 π π and thus π = π since STOCHASTIC CALCULUS ITO INTEGRAL ο π π 0 Let π π‘ ππ π‘ = lim π→∞ ππ π ππ π−1 π π=0 π = ππ and note that 1 2 π π+1 π π π−1 π=0 (ππ+1 −π = = = π−1 π=0 ππ 1 2 ππ+1 − ππ = ππ2 − 1 2 1 π−1 1 π−1 π−1 2 2 π − π π + π π π+1 π=0 2 π=0 π+1 2 π=0 π+1 1 π 1 π−1 π−1 2 2 π − π π + ππ+1 π π+1 π=1 π π=0 π=0 2 2 1 1 π−1 1 π−1 π−1 2 2 ππ2 + π − π π + ππ+1 π π+1 π=0 π π=0 π=0 2 2 2 1 ππ2 + π−1 ππ2 − π−1 π=0 π=0 ππ ππ+1 2 1 ππ2 + π−1 π=0 ππ ππ − ππ+1 2 − ππ )2 = = Thus ππ π π−1 π=0 (ππ+1 − ππ )2 ITO INTEGRAL ο π π 0 π‘ ππ π‘ = lim π→∞ = ππ π−1 π π=0 π 1 lim π 2 π→∞ 2 1 2 = π2 π − π − 1 2 1 2 π π−1 π=0 π+1 π π π −π π+1 π π ππ π −π ππ π 2 π, π π Quadratic Variation 1 2 = π2 π − 1 π 2 ITO’S LEMMA ο We seek an approximation π π, π π − π 0, π 0 By Taylor’s formula we have π π‘π+1 , π₯π+1 − π π‘π , π₯π = ππ‘ π‘π , π₯π π‘π+1 − π‘π + ππ₯ π‘π , π₯π π₯π+1 ITO’S LEMMA ο π π, π π + + − π 0, π 0 π−1 π=0 ππ₯ π‘π , π π‘π π−1 π=0 ππ₯π‘ π‘π , π π‘π = π−1 π=0 π π‘π+1 , π π‘π+1 π π‘π+1 − π(π‘π ) + 1 2 − π π‘π , π π‘π π−1 π=0 ππ₯π₯ π‘π+1 − π‘π π π‘π+1 − π(π‘π ) + π‘π , π π‘π 1 2 π−1 π=0 ππ‘π‘ = π−1 π=0 ππ‘ π‘π , π π‘π π(π‘π+1 ) − π(π‘π ) π‘π , π π‘π 2 π‘π+1 − π‘π π‘π+1 ) − π(π‘π ) 2 ο Now taking limits, lim π π, π π − π 0, π 0 = π π, π π − π 0, π 0 π →∞ π π π = 0 ππ‘ π‘, π(π‘) ππ‘ + 0 ππ₯ π‘, π π‘ ππ π‘ + 0 ππ₯π₯ π‘, π π‘ ππ‘ since ππ‘ππ π‘ = 0 and ππ‘ππ‘ = 0 ο In differential form, Ito’s formula is ππ(π‘, π π‘ = ππ‘ π‘, π π‘ ππ‘ + ππ₯ π‘, π π‘ ππ π‘ 1 1 + 2 ππ₯π₯ π‘, π π‘ ππ π‘ ππ π‘ + ππ₯π‘ π‘, π π‘ ππ π‘ ππ‘ + 2 ππ‘π‘ π‘, π π‘ ππ‘ππ‘ with the last two terms cancelling out to zero ITO’S LEMMA ο Ex) Suppose π π‘, π₯ = π‘π₯ 2 . What is ππ(π‘, π π‘ )? ππ‘ π‘, π₯ = π₯ 2 ππ₯ π‘, π₯ = 2π‘π₯ ππ₯π₯ π‘, π₯ = 2π‘ Then ππ π‘, π π‘ = π 2 π‘ ππ‘ + 2π‘π π‘ ππ π‘ + 1 2π‘ 2 ππ π‘ 2 = πΎπ (π) ITO’S LEMMA ο Ex) Suppose π π‘, π₯ = sin π‘π₯ . What is ππ(π‘, π π‘ )? ππ‘ π‘, π₯ = π₯ ∗ cos π‘π₯ ππ₯ π‘, π₯ = π‘ ∗ cos(π‘π₯) ππ₯π₯ π‘, π₯ = −π‘ 2 ∗ sin π‘π₯ Then ππ π‘, π π‘ = (π₯ ∗ cos π‘π₯ )ππ‘ + π‘ ∗ cos π‘π₯ ππ π‘ + ∗ ππ¨π¬ ππ − ππ ∗ π¬π’π§ ππ )π π + π ∗ ππ¨π¬ ππ π πΎ π −π‘ 2 ∗ sin π‘π₯ ππ π‘ 2 = (π ITO’S LEMMA ο More generally, if π π‘ is a stochastic process ππ π‘, π π‘ 1 = ππ‘ π‘, π π‘ ππ‘ + ππ₯ π‘, π π‘ ππ π‘ + ππ₯π₯ π‘, π π‘ 2 ππ π‘ 2 We have been using Ito’s formula to construct stochastic differential equations (SDE’s)—that is, differential equations with a random term. Consider the SDE: ππ π‘ = π π‘ ππ π‘ + πΌ π‘ − If π π‘ = π 0 exp π π‘ , what is ππ(π‘)? 1 2 π (π‘) 2 ππ‘ ITO’S LEMMA ο Here, π π‘ = π π‘, π₯ = π 0 exp π₯ ο Note that this is actually just a function of a single variable x ππ‘ π‘, π₯ = 0 ππ₯ π‘, π₯ = π 0 exp{π₯} ππ₯π₯ π‘, π₯ = π 0 exp{π₯} Then ππ π‘ = π 0 π π π‘ ππ π‘ + 1 π 2 0 ππ π‘ ππ π‘ 2 ITO’S LEMMA ο Note that ππ π‘ 2 = π π‘ ππ π‘ + πΌ π‘ − 1 2 π 2 π‘ ππ‘ 2 = π2 π‘ ππ π‘ 2 + 2π π‘ πΌ π‘ BLACK-SCHOLES EQUATION BLACK-SCHOLES ο Let the underlying follow this SDE with constant rate and volatility: ππ π‘ = πΌπ π‘ ππ‘ + ππ π‘ ππ π‘ ο The only variable inputs to an options price are the time until maturity and the price of the stock, so we start by considering the function π(π‘, π(π‘)) Ito’s formula tells us ππ π‘, π π‘ = ππ‘ π‘, π π‘ ππ‘ + ππ₯ π‘, π π‘ ππ π‘ + 1 1 π 2 π₯π₯ π‘, π π‘ ππ π‘ ππ π‘ = ππ‘ π‘, π π‘ ππ‘ + ππ₯ π‘, π π‘ (πΌπ π‘ ππ‘ + ππ π‘ ππ π‘ ) + ππ₯π₯ π‘, π π‘ π 2 π 2 π‘ ππ‘ 2 1 2 = ππ‘ π‘, π π‘ + πΌπ π‘ ππ₯ π‘, π π‘ + π π‘ π 2 π‘ ππ₯π₯ π‘, π π‘ ππ‘ + ππ(π‘)ππ₯ π‘, π π‘ ππ(π‘) 2 BLACK SCHOLES ο We need to take the present value of this so we consider the function: π −ππ‘ π(π‘, π(π‘)) = −ππ −ππ‘ π π‘, π π‘ ππ‘ + π −ππ‘ ππ π‘, π π‘ 1 = π −ππ‘ −ππ π‘, π π‘ + ππ‘ π‘, π π‘ + πΌπ π‘ ππ₯ π‘, π π‘ + π 2 π‘ π 2 π‘ ππ₯π₯ π‘, π π‘ 2 −ππ‘ + π ππ(π‘)ππ₯ π‘, π π‘ ππ(π‘) Again, by Ito’s formula ππ π‘, π π‘ ππ‘ BLACK SCHOLES ο Meanwhile, we try to replicate the option contract as we did in the binomial option pricing model. That is, by investing some money in a stock position and some in some money market account (a bond): ο Let π π‘ be the value of our portfolio at time π‘ ο At time π‘ we invest a necessary amount β π‘ into the stock and the remainder, π π‘ − β(π‘), into the money market instrument. ο Then we gain β π‘ ππ(π‘) from our investment in the stock ο And π π π‘ − β π‘ ππ‘ from our investment in the money market instrument ο Thus ππ π‘ = β π‘ ππ π‘ + π π π‘ − β π‘ ππ‘ = β π‘ πΌπ π‘ ππ‘ + ππ π‘ ππ π‘ ο By Ito’s lemma, the differential of the PV(stock) is π π −ππ‘ π π‘ + π π π‘ − β π‘ π π‘ ππ‘ = πΌ − π π −ππ‘ π π‘ ππ‘ + ππ −ππ‘ π π‘ ππ π‘ ο Likewise, the differential of our discounted portfolio is π π −ππ‘ π π‘ + β π‘ ππ −ππ‘ π π‘ ππ π‘ = β π‘ πΌ − π π −ππ‘ π π‘ ππ‘ BLACK SCHOLES ο At each time π‘, we want the replicating portfolio π π‘ to match the value of the option π π‘, π π‘ ο We do this by ensuring that π π −ππ‘ π π‘ = π π −ππ‘ π π‘, π π‘ β π‘ πΌ − π π −ππ‘ π π‘ ππ‘ + β π‘ ππ −ππ‘ π π‘ ππ π‘ = π −ππ‘ −ππ π‘, π π‘ + ππ‘ π‘, π π‘ + π −ππ‘ ππ(π‘)ππ₯ π‘, π π‘ ππ(π‘) + ππ π‘ ππ₯ π‘, π π‘ + for all π‘ and that π 0 = π 0, π 0 : 1 2 π π‘ π 2 π‘ ππ₯π₯ π‘, π π‘ 2 ππ‘ BLACK SCHOLES ο At each time π‘, we want the replicating portfolio π π‘ to match the value of the option π π‘, π π‘ ο We do this by ensuring that π π −ππ‘ π π‘ = π π −ππ‘ π π‘, π π‘ β π‘ πΌ − π π π‘ ππ‘ + β π‘ ππ π‘ ππ π‘ = −ππ π‘, π π‘ + ππ‘ π‘, π π‘ + πΌπ π‘ ππ₯ π‘, π π‘ + for all π‘ and that π 0 = π 0, π 0 : 1 2 π π‘ π 2 π‘ ππ₯π₯ π‘, π π‘ 2 ππ‘ + ππ(π‘)ππ₯ π‘, π π‘ ππ(π‘) BLACK-SCHOLES ο At each time π‘, we want the replicating portfolio π π‘ to match the value of the option π π‘, π π‘ ο We do this by ensuring that π π −ππ‘ π π‘ = π π −ππ‘ π π‘, π π‘ β π‘ πΌ − π π π‘ ππ‘ + β π ππΊ π π πΎ π = −ππ π‘, π π‘ + ππ‘ π‘, π π‘ + πΌπ π‘ ππ₯ π‘, π π‘ Need β π ππΊ π = ππΊ π ππ π, πΊ π + for all π‘ and that π 0 = π 0, π 0 : 1 2 π π‘ π 2 π‘ ππ₯π₯ π‘, π π‘ 2 → β π = ππ π, πΊ π ππ‘ + ππΊ(π)ππ π, πΊ π π πΎ(π) BLACK-SCHOLES ο At each time π‘, we want the replicating portfolio π π‘ to match the value of the option π π‘, π π‘ ο We do this by ensuring that π π −ππ‘ π π‘ = π π −ππ‘ π π‘, π π‘ β π πΆ − π πΊ π π π + β π‘ ππ π‘ ππ π‘ = −ππ π, πΊ π + ππ π, πΊ π + πΆπΊ π ππ π, πΊ π Need β π‘ ππ π‘ = ππ π‘ ππ₯ π‘, π π‘ + for all π‘ and that π 0 = π 0, π 0 : π π π π πΊπ π πππ π, πΊ π π π π + ππ(π‘)ππ₯ π‘, π π‘ ππ(π‘) → β π‘ = ππ₯ π‘, π π‘ 1 Needβ π‘ πΌ − π π π‘ = −ππ π‘, π π‘ + ππ‘ π‘, π π‘ + πΌπ π‘ ππ₯ π‘, π π‘ + 2 π 2 π‘ π 2 π‘ ππ₯π₯ π‘, π π‘ BLACK-SCHOLES ο At each time π‘, we want the replicating portfolio π π‘ to match the value of the option π π‘, π π‘ ο We do this by ensuring that π π −ππ‘ π π‘ = π π −ππ‘ π π‘, π π‘ β π πΆ − π πΊ π π π + β π‘ ππ π‘ ππ π‘ = −ππ π, πΊ π + ππ π, πΊ π + πΆπΊ π ππ π, πΊ π Need β π‘ ππ π‘ = ππ π‘ ππ₯ π‘, π π‘ Need ππ₯ π‘, π π‘ π π π π πΊπ π πππ π, πΊ π π π π + ππ(π‘)ππ₯ π‘, π π‘ ππ(π‘) → β π‘ = ππ₯ π‘, π π‘ πΌ − π π π‘ = −ππ π‘, π π‘ Simplifying this we need, ππ π‘, π π‘ + for all π‘ and that π 0 = π 0, π 0 : + ππ‘ π‘, π π‘ = ππ‘ π‘, π π‘ + πΌπ π‘ ππ₯ π‘, π π‘ + ππ π‘ ππ₯ π‘, π π‘ 1 + 1 2 π 2 π‘ π 2 π‘ ππ₯π₯ π‘, π π‘ + 2 π 2 π‘ π 2 π‘ ππ₯π₯ π‘, π π‘ BLACK-SCHOLES ππ π‘, π π‘ With = ππ‘ π‘, π π‘ + ππ π‘ ππ₯ π‘, π π‘ 1 + π 2 π 2 π‘ ππ₯π₯ π‘, π π‘ 2 ∀π‘ ∈ 0, π , π₯ ≥ 0 π π, π₯ = max{π₯ − πΎ, 0} Is the Black-Scholes-Merton partial differential equation. Its is a backward parabolic equation, which are known to have solutions. Using the fact that, ππ‘ π‘, 0 = ππ π‘, 0 , we solve this ODE: π π‘, 0 = π −ππ‘ π 0,0 = π −ππ‘ ∗ 0 = 0. This gives us our first boundary condition at π₯ = 0: π π‘, 0 = 0, ∀π‘ Additionally, lim π π‘, π₯ − π₯ − π −π π₯→∞ π−π‘ πΎ = 0, ∀π‘ ∈ 0, π That is, the fact that as the underlying approaches ∞, the call option begins to look like the underlying minus the discounted strike. This serves as the second boundary condition. BLACK-SCHOLES ο Solving the Black-Scholes-Merton PDE gives us the familiar results: π, π‘ = π π1 π − π π2 πΎπ −π(π−π‘) π π, π‘ = π −π2 πΎπ −π(π−π‘) − π −π1 π 1 π π1 = ∗ ππ + πΎ π π−π‘ 1 π π2 = ∗ ππ + πΎ π π−π‘ π2 π+ 2 π2 π− 2 π−π‘ π−π‘ π π₯ is the standard-normal CDF of x BLACK-SCHOLES ο Why doesn’t this method work for American options? ο Early exercise is not modeled! ο Pros ο Gives an analytical (no algorithms necessary!) solution to the value of a European option ο This is simple enough to be extended ο The resulting PDE’s can be solved numerically ο Cons ο Some unrealistic assumptions about rates and volatilitiesο does not match data ο Normal distribution has thin talesο under-approximates large returns in stocks THANK YOU! ο Facebook: http://www.facebook.com/UIUCFEC ο LinkedIn: http://www.linkedin.com/financialengineeringclub ο Email: uiuc.fec@gmail.com President Greg Pastorek gfpastorek@gmail.com Internal Vice President Matthew Reardon mreardon5@gmail.com