You are on page 1of 365

Steven Shreve: Stochastic Calculus and Finance

P RASAD C HALASANI Carnegie Mellon University chal@cs.cmu.edu S OMESH J HA Carnegie Mellon University sjha@cs.cmu.edu

THIS IS A DRAFT: PLEASE DO NOT DISTRIBUTE c Copyright; Steven E. Shreve, 1996

October 6, 1997

Contents
1 Introduction to Probability Theory 1.1 1.2 1.3 1.4 1.5 The Binomial Asset Pricing Model . . . . . . . . . . . . . . . . . . . . . . . . . . Finite Probability Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lebesgue Measure and the Lebesgue Integral . . . . . . . . . . . . . . . . . . . . General Probability Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Independence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.1 1.5.2 1.5.3 1.5.4 1.5.5 1.5.6 1.5.7 Independence of sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Independence of -algebras . . . . . . . . . . . . . . . . . . . . . . . . . Independence of random variables . . . . . . . . . . . . . . . . . . . . . . Correlation and independence . . . . . . . . . . . . . . . . . . . . . . . . Independence and conditional expectation. . . . . . . . . . . . . . . . . . Law of Large Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . Central Limit Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 11 16 22 30 40 40 41 42 44 45 46 47 49 49 50 52 52 53 54 55 57 58

2 Conditional Expectation 2.1 2.2 2.3 A Binomial Model for Stock Price Dynamics . . . . . . . . . . . . . . . . . . . . Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conditional Expectation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1 2.3.2 2.3.3 2.3.4 2.3.5 2.4 An example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Denition of Conditional Expectation . . . . . . . . . . . . . . . . . . . . Further discussion of Partial Averaging . . . . . . . . . . . . . . . . . . . Properties of Conditional Expectation . . . . . . . . . . . . . . . . . . . . Examples from the Binomial Model . . . . . . . . . . . . . . . . . . . . .

Martingales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 3 Arbitrage Pricing 3.1 3.2 3.3 Binomial Pricing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . General one-step APT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Risk-Neutral Probability Measure . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 3.3.2 3.4 3.5 Portfolio Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Self-nancing Value of a Portfolio Process . . . . . . . . . . . . . . . . 59 59 60 61 62 62 63 64 67 67 69 70 70 73 74 77 77 79 81 85 85 86 88 89 91 91 92 94 97 97

Simple European Derivative Securities . . . . . . . . . . . . . . . . . . . . . . . . The Binomial Model is Complete . . . . . . . . . . . . . . . . . . . . . . . . . . .

4 The Markov Property 4.1 4.2 4.3 4.4 4.5 Binomial Model Pricing and Hedging . . . . . . . . . . . . . . . . . . . . . . . . Computational Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Markov Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Different ways to write the Markov property . . . . . . . . . . . . . . . . Showing that a process is Markov . . . . . . . . . . . . . . . . . . . . . . . . . . Application to Exotic Options . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5 Stopping Times and American Options 5.1 5.2 5.3 American Pricing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Value of Portfolio Hedging an American Option . . . . . . . . . . . . . . . . . . . Information up to a Stopping Time . . . . . . . . . . . . . . . . . . . . . . . . . .

6 Properties of American Derivative Securities 6.1 6.2 6.3 6.4 The properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Proofs of the Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Compound European Derivative Securities . . . . . . . . . . . . . . . . . . . . . . Optimal Exercise of American Derivative Security . . . . . . . . . . . . . . . . . .

7 Jensens Inequality 7.1 7.2 7.3 Jensens Inequality for Conditional Expectations . . . . . . . . . . . . . . . . . . . Optimal Exercise of an American Call . . . . . . . . . . . . . . . . . . . . . . . . Stopped Martingales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

8 Random Walks 8.1 First Passage Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

8.5 8.6 8.7 8.8 8.9

The Strong Markov Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 General First Passage Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 Example: Perpetual American Put . . . . . . . . . . . . . . . . . . . . . . . . . . 102 Difference Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 Distribution of First Passage Times . . . . . . . . . . . . . . . . . . . . . . . . . . 107

8.10 The Reection Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 9 Pricing in terms of Market Probabilities: The Radon-Nikodym Theorem. 9.1 9.2 9.3 9.4 9.5 111

Radon-Nikodym Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 Radon-Nikodym Martingales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 The State Price Density Process . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 Stochastic Volatility Binomial Model . . . . . . . . . . . . . . . . . . . . . . . . . 116 Another Applicaton of the Radon-Nikodym Theorem . . . . . . . . . . . . . . . . 118 119

10 Capital Asset Pricing

10.1 An Optimization Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 11 General Random Variables 123

11.1 Law of a Random Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 11.2 Density of a Random Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 11.3 Expectation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 11.4 Two random variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 11.5 Marginal Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 11.6 Conditional Expectation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 11.7 Conditional Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 11.8 Multivariate Normal Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 11.9 Bivariate normal distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 11.10MGF of jointly normal random variables . . . . . . . . . . . . . . . . . . . . . . . 130 12 Semi-Continuous Models 131

12.1 Discrete-time Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

8.4

Expectation of

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

8.3

The moment generating function for

8.2

is almost surely nite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

97 99

4 12.2 The Stock Price Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132 12.3 Remainder of the Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 12.4 Risk-Neutral Measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 12.5 Risk-Neutral Pricing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 12.6 Arbitrage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 12.7 Stalking the Risk-Neutral Measure . . . . . . . . . . . . . . . . . . . . . . . . . . 135 12.8 Pricing a European Call . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 13 Brownian Motion 139

13.1 Symmetric Random Walk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 13.2 The Law of Large Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 13.3 Central Limit Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 13.4 Brownian Motion as a Limit of Random Walks . . . . . . . . . . . . . . . . . . . 141 13.5 Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 13.6 Covariance of Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 13.7 Finite-Dimensional Distributions of Brownian Motion . . . . . . . . . . . . . . . . 144 13.8 Filtration generated by a Brownian Motion . . . . . . . . . . . . . . . . . . . . . . 144 13.9 Martingale Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 13.10The Limit of a Binomial Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 13.11Starting at Points Other Than 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 13.12Markov Property for Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . 147 13.13Transition Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 13.14First Passage Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 14 The It Integral o 153

14.1 Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 14.2 First Variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 14.3 Quadratic Variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155 14.4 Quadratic Variation as Absolute Volatility . . . . . . . . . . . . . . . . . . . . . . 157 14.5 Construction of the It Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 o 14.6 It integral of an elementary integrand . . . . . . . . . . . . . . . . . . . . . . . . 158 o 14.7 Properties of the It integral of an elementary process . . . . . . . . . . . . . . . . 159 o 14.8 It integral of a general integrand . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 o

5 14.9 Properties of the (general) It integral . . . . . . . . . . . . . . . . . . . . . . . . 163 o 14.10Quadratic variation of an It integral . . . . . . . . . . . . . . . . . . . . . . . . . 165 o 15 It s Formula o 167

15.1 It s formula for one Brownian motion . . . . . . . . . . . . . . . . . . . . . . . . 167 o 15.2 Derivation of It s formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 o 15.3 Geometric Brownian motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 15.4 Quadratic variation of geometric Brownian motion . . . . . . . . . . . . . . . . . 170 15.5 Volatility of Geometric Brownian motion . . . . . . . . . . . . . . . . . . . . . . 170 15.6 First derivation of the Black-Scholes formula . . . . . . . . . . . . . . . . . . . . 170 15.7 Mean and variance of the Cox-Ingersoll-Ross process . . . . . . . . . . . . . . . . 172 15.8 Multidimensional Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . . 173 15.9 Cross-variations of Brownian motions . . . . . . . . . . . . . . . . . . . . . . . . 174 15.10Multi-dimensional It formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 o 16 Markov processes and the Kolmogorov equations 177

16.1 Stochastic Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 16.2 Markov Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178 16.3 Transition density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179 16.4 The Kolmogorov Backward Equation . . . . . . . . . . . . . . . . . . . . . . . . 180 16.5 Connection between stochastic calculus and KBE . . . . . . . . . . . . . . . . . . 181 16.6 Black-Scholes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 16.7 Black-Scholes with price-dependent volatility . . . . . . . . . . . . . . . . . . . . 186 17 Girsanovs theorem and the risk-neutral measure 189

17.2 Risk-neutral measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193 18 Martingale Representation Theorem 197

18.1 Martingale Representation Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 197 18.2 A hedging application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 18.4 18.3 -dimensional Girsanov Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 199

-dimensional Martingale Representation Theorem . . . . . . . . . . . . . . . . . 200

18.5 Multi-dimensional market model . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

17.1 Conditional expectations under

. . . . . . . . . . . . . . . . . . . . . . . . . . 191

6 19 A two-dimensional market model 19.2 Hedging when 203

20 Pricing Exotic Options

20.1 Reection principle for Brownian motion . . . . . . . . . . . . . . . . . . . . . . 209 20.2 Up and out European call. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212 20.3 A practical issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218 21 Asian Options 219

21.1 Feynman-Kac Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220 21.2 Constructing the hedge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220 21.3 Partial average payoff Asian option . . . . . . . . . . . . . . . . . . . . . . . . . . 221 22 Summary of Arbitrage Pricing Theory 223

22.1 Binomial model, Hedging Portfolio . . . . . . . . . . . . . . . . . . . . . . . . . 223 22.2 Setting up the continuous model . . . . . . . . . . . . . . . . . . . . . . . . . . . 225 22.3 Risk-neutral pricing and hedging . . . . . . . . . . . . . . . . . . . . . . . . . . . 227 22.4 Implementation of risk-neutral pricing and hedging . . . . . . . . . . . . . . . . . 229 23 Recognizing a Brownian Motion 233

23.1 Identifying volatility and correlation . . . . . . . . . . . . . . . . . . . . . . . . . 235 23.2 Reversing the process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236 24 An outside barrier option 239

24.1 Computing the option value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242 24.2 The PDE for the outside barrier option . . . . . . . . . . . . . . . . . . . . . . . . 243 24.3 The hedge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245 25 American Options 247

25.1 Preview of perpetual American put . . . . . . . . . . . . . . . . . . . . . . . . . . 247 25.2 First passage times for Brownian motion: rst method . . . . . . . . . . . . . . . . 247 25.3 Drift adjustment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249 25.4 Drift-adjusted Laplace transform . . . . . . . . . . . . . . . . . . . . . . . . . . . 250 25.5 First passage times: Second method . . . . . . . . . . . . . . . . . . . . . . . . . 251

     

19.1 Hedging when

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 209

7 25.6 Perpetual American put . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252 25.7 Value of the perpetual American put . . . . . . . . . . . . . . . . . . . . . . . . . 256 25.8 Hedging the put . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257 25.9 Perpetual American contingent claim . . . . . . . . . . . . . . . . . . . . . . . . . 259 25.10Perpetual American call . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259 25.11Put with expiration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260 25.12American contingent claim with expiration . . . . . . . . . . . . . . . . . . . . . 261 26 Options on dividend-paying stocks 263

26.1 American option with convex payoff function . . . . . . . . . . . . . . . . . . . . 263 26.2 Dividend paying stock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264 26.3 Hedging at time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266 267

27 Bonds, forward contracts and futures

27.1 Forward contracts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269 27.2 Hedging a forward contract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269 27.3 Future contracts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270 27.4 Cash ow from a future contract . . . . . . . . . . . . . . . . . . . . . . . . . . . 272 27.5 Forward-future spread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272 27.6 Backwardation and contango . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273 28 Term-structure models 275

28.1 Computing arbitrage-free bond prices: rst method . . . . . . . . . . . . . . . . . 276 28.2 Some interest-rate dependent assets . . . . . . . . . . . . . . . . . . . . . . . . . 276 28.3 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277 28.4 Forward rate agreement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277 28.6 Computing arbitrage-free bond prices: Heath-Jarrow-Morton method . . . . . . . . 279 28.7 Checking for absence of arbitrage . . . . . . . . . . . . . . . . . . . . . . . . . . 280 28.8 Implementation of the Heath-Jarrow-Morton model . . . . . . . . . . . . . . . . . 281 29 Gaussian processes 285 28.5 Recovering the interest from the forward rate . . . . . . . . . . . . . . . . . . 278

29.1 An example: Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286 30 Hull and White model 293

' $ (&%#

! "

8 30.1 Fiddling with the formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295 30.2 Dynamics of the bond price . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296 30.3 Calibration of the Hull & White model . . . . . . . . . . . . . . . . . . . . . . . . 297 30.4 Option on a bond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299 31 Cox-Ingersoll-Ross model 303 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306

31.2 Kolmogorov forward equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306 31.3 Cox-Ingersoll-Ross equilibrium density . . . . . . . . . . . . . . . . . . . . . . . 309 31.4 Bond prices in the CIR model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310 31.5 Option on a bond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313 31.6 Deterministic time change of CIR model . . . . . . . . . . . . . . . . . . . . . . . 313 31.7 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315 31.8 Tracking down in the time change of the CIR model . . . . . . . . . . . . . 316 319

32 A two-factor model (Dufe & Kan)

32.2 Zero-coupon bond prices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321 32.3 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323 33 Change of num raire e 325

33.1 Bond price as num raire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327 e 33.2 Stock price as num raire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328 e 33.3 Merton option pricing formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329 34 Brace-Gatarek-Musiela model 335 . . . . . . . . . . . . . . . . . . . . . . . . . 335

34.2 Brace-Gatarek-Musiela model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336 34.3 LIBOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337 34.4 Forward LIBOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338 34.6 Implementation of BGM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340 34.7 Bond prices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342 34.8 Forward LIBOR under more forward measure . . . . . . . . . . . . . . . . . . . . 343 34.5 The dynamics of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338

34.1 Review of HJM under risk-neutral

' A $ )BB&@9

32.1 Non-negativity of

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320

' $ 10)#

31.1 Equilibrium distribution of

' 6$3 75542

9 34.9 Pricing an interest rate caplet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343 34.10Pricing an interest rate cap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345 34.11Calibration of BGM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345 34.12Long rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346 34.13Pricing a swap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346 35 Notes and References 349

35.1 Probability theory and martingales. . . . . . . . . . . . . . . . . . . . . . . . . . . 349 35.2 Binomial asset pricing model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349 35.3 Brownian motion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350 35.4 Stochastic integrals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350 35.5 Stochastic calculus and nancial markets. . . . . . . . . . . . . . . . . . . . . . . 350 35.6 Markov processes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351 35.7 Girsanovs theorem, the martingale representation theorem, and risk-neutral measures.351 35.8 Exotic options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352 35.9 American options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352 35.10Forward and futures contracts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353 35.11Term structure models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353 35.12Change of num raire. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353 e 35.13Foreign exchange models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353 35.14REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354

10

Chapter 1

Introduction to Probability Theory


1.1 The Binomial Asset Pricing Model
The binomial asset pricing model provides a powerful tool to understand arbitrage pricing theory and probability theory. In this course, we shall use it for both these purposes. In the binomial asset pricing model, we model stock prices in discrete time, assuming that at each step, the stock price will change to one of two possible values. Let us begin with an initial positive stock price . There are two positive numbers, and , with

such that at the next period, the stock price will be either or . Typically, we take and to satisfy , so change of the stock price from to represents a downward movement, and change of the stock price from to represents an upward movement. It is common to also have , and this will be the case in many of our examples. However, strictly speaking, for what we are about to do we need to assume only (1.1) and (1.2) below. Of course, stock price movements are much more complicated than indicated by the binomial asset pricing model. We consider this simple model for three reasons. First of all, within this model the concept of arbitrage pricing and its relation to risk-neutral pricing is clearly illuminated. Secondly, the model is used in practice because with a sufcient number of steps, it provides a good, computationally tractable approximation to continuous-time models. Thirdly, within the binomial model we can develop the theory of conditional expectations and martingales which lies at the heart of continuous-time models. With this third motivation in mind, we develop notation for the binomial model which is a bit different from that normally found in practice. Let us imagine that we are tossing a coin, and when we get a Head, the stock price moves up, but when we get a Tail, the price moves down. We denote the price at time by if the toss results in head (H), and by if it 11

D C  ' e$ ! EX4@0b4C

DC EX

DC ESF

D EC

DC RQ

C D YF

AF   PIHG6 DC D C F  ' c$ ! ES4d5b4C  ! a ` F    UWVTUT6

D EC

(1.1)

12
S2 (HH) = 16

S (H) = 8 1 S2 (HT) = 4 S =4 0 S1 (T) = 2 S2 (TH) = 4

S2 (TT) = 1

results in tail (T). After the second toss, the price will be one of:

After three tosses, there are eight possible coin sequences, although not all of them result in different stock prices at time . For the moment, let us assume that the third toss is the last one and denote by

the set of all possible outcomes of the three tosses. The set of all possible outcomes of a random experiment is called the sample space for the experiment, and the elements of are called sample points. In this case, each sample point is a sequence of length three. We denote the -th component of by . For example, when , we have , and . at time depends on the coin tosses. To emphasize this, we often write . The stock price Actually, this notation does not quite tell the whole story, for while depends on all of , depends on only the rst two components of , depends on only the rst component of , and does not depend on at all. Sometimes we will use notation such just to record more explicitly how depends on . Example 1.1 Set , and . We have then the binomial tree of possible stock prices shown in Fig. 1.1. Each sample point represents a path through the tree. Thus, we can think of the sample space as either the set of all possible outcomes from three coin tosses or as the set of all possible paths through the tree. To complete our binomial asset pricing model, we introduce a money market with interest rate ; $1 invested in the money market becomes in the next period. We take to be the interest

q PC ' d$ C c  @r

p  h  f  D 7iF g@EC e  q c  ! x  # ' q eA ! d$ q C EC

Figure 1.1: Binomial tree of stock prices with

e e eA c e eA e c eA c c eA e e cA c e cA e c cA c c c 1w@1w@1wd1@Bww"wdBwdu

D  ' e$ ! C  ' e e$ q A D C F  ' e$ ! C F  ' c e$ q 7EC q t@0s4Qt@07PC uEXSt@&b4S4d@07PC A D C F  ' c$ ! C  ' e c$ q C A D F  ' c$ ! C F  ' c c$ q bRSXtd5yxQ4@w57PvuEC q td5s4Yrdd57PC ' A ireAuqx1l!d$k  p  f vWF g D C !q j ' Aq A! $  sresxeihdg q fC c e c  @w '# n $ uHh1Sm ! xC t C DC

CHAPTER 1. Introduction to Probability Theory

13

rate for both borrowing and lending. (This is not as ridiculous as it rst seems, because in a many applications of the model, an agent is either borrowing or lending (not both) and knows in advance which she will be doing; in such an application, she should take to be the rate of interest for her activity.) We assume that

The model would not make sense if we did not have this condition. For example, if , then the rate of return on the money market is always at least as great as and sometimes greater than the return on the stock, and no one would invest in the stock. The inequality cannot happen unless either is negative (which never happens, except maybe once upon a time in Switzerland) or . In the latter case, the stock does not really go down if we get a tail; it just goes up less than if we had gotten a head. One should borrow money at interest rate and invest in the stock, since even in the worst case, the stock price rises at least as fast as the debt used to buy it. With the stock as the underlying asset, let us consider a European call option with strike price and expiration time . This option confers the right to buy the stock at time for dollars, and so is worth at time if is positive and is otherwise worth zero. We denote by

the value (payoff) of this option at expiration. Of course, actually depends only on , and we can and do sometimes write rather than . Our rst task is to compute the arbitrage price of this option at time zero. Suppose at time zero you sell the call for dollars, where is still to be determined. You now have an obligation to pay off if and to pay off if . At the time you sell the option, you dont yet know which value will take. You hedge your short shares of stock, where is still to be determined. You can use position in the option by buying the proceeds of the sale of the option for this purpose, and then borrow if necessary at interest rate to complete the purchase. If is more than necessary to buy the shares of stock, you dollars invested invest the residual money at interest rate . In either case, you will have in the money market, where this quantity might be negative. You will also own shares of stock. If the stock goes up, the value of your portfolio (excluding the short position in the option) is

and you need to have

. Thus, you want to choose

and

so that (1.3)

If the stock goes down, the value of your portfolio is

and you need to have

. Thus, you want to choose

and

to also have (1.4)

! h

e  ! ~

F r # n Hpq4 s

# n r I`

D ! v' c  ! v~ s @DECX5$ Du ' $ ds! u ' $ }y! u 6 ' $ ! C { z y x v XBA s |dyxb)uTw'

DG D C D D D G u

llDEsDGWD u suT1Ed5b4lGtd5s! u ' C $ ' # n $ n ' c$ ! C D  ' c$ D Du ' c$ d5s! u A l' D C D W D u suHt1E$ ! C D $'# n $ n ' c

lsDElDGD u suHt10b4lGt@0s! u ' C $ ' # n $ n ' e$ ! C D  ' e$ ' e @0$ ! u D Du A ' D C D $ ' # n $ n ' e$ ! C D llEsGWD u suT1E@&b4sG

(1.2)

F  # n  fHpHo ' $ ! C$  ' $ s 4dyx5gtds! u Du ! s 4C v ' s @EY5$ DCF # '! $ e}b! u Du D ! s 4C Du # 6 ks t # r vW

14 These are two equations in two unknowns, and we solve them below Subtracting (1.4) from (1.3), we obtain

so that

This is a discrete-time version of the famous delta-hedging formula for derivative securities, according to which the number of shares of an underlying asset a hedge should hold is the derivative (in the sense of calculus) of the value of the derivative security with respect to the price of the underlying asset. This formula is so pervasive the when a practitioner says delta, she means the derivative (in the sense of calculus) just described. Note, however, that my denition of is the number of shares of stock one holds at time zero, and (1.6) is a consequence of this denition, not the denition of itself. Depending on how uncertainty enters the model, there can be cases in which the number of shares of stock a hedge should hold is not the (calculus) derivative of the derivative security with respect to the price of the underlying asset. To complete the solution of (1.3) and (1.4), we substitute (1.6) into either (1.3) or (1.4) and solve for . After some simplication, this leads to the formula

This is the arbitrage price for the European call option with payoff formula, we dene

at time . To simplify this

so that (1.7) becomes

Because we have taken , both and are dened,i.e., the denominator in (1.8) is not zero. , and because they sum to , we can regard Because of (1.2), both and are in the interval them as probabilities of and , respectively. They are the risk-neutral probabilites. They appeared when we solved the two equations (1.3) and (1.4), and have nothing to do with the actual probabilities of getting or on the coin tosses. In fact, at this point, they are nothing more than a convenient tool for writing (1.7) as (1.9). We now consider a European call which pays off dollars at time . At expiration, the payoff of this option is , where and depend on and , the rst and second coin tosses. We want to determine the arbitrage price for this option at time zero. Suppose an agent sells the option at time zero for dollars, where is still to be determined. She then buys shares

D G

e c 'A6 ss)$ F  gW ' e$ (0y! u Ed5s! u r I t D u n ' c$ # n   A  uTWeoF F x  A WqHF x  ' # H n $ # n ! u # n '@0b! ' # q oF n ' c $ W# F n T  e$ u uHt1F Ed5y! u WH D u n $

A'' e C ' c C  ' e l"@0$ ! Wrd5$ ! 5$ D t@0$ ! u rd5$ ! u ' c p ! ' C 4d5$ ' c @0&$ee y!! q4d5s!! C G ' @$ u ' c$ u  D s qC Du qu Du v ' s q 5Vx  q u C$ e c D Du

(1.5)

(1.6)

(1.7)

(1.8)

(1.9)

CHAPTER 1. Introduction to Probability Theory

15

dollars in the money market to nance this. At time , the agent has of stock, investing a portfolio (excluding the short position in the option) valued at

After the rst coin toss, the agent has dollars and can readjust her hedge. Suppose she decides to now hold shares of stock, where is allowed to depend on because the agent knows what value has taken. She invests the remainder of her wealth, in the money market. In the next period, her wealth will be given by the right-hand side of the following equation, and she wants it to be . Therefore, she wants to have

Although we do not indicate it in the notation, and depend on and , the outcomes of the rst two coin tosses. Considering all four possible outcomes, we can write (1.11) as four equations:

We now have six equations, the two represented by (1.10) and the four represented by (1.11), in the six unknowns , , , , , and . To solve these equations, and thereby determine the arbitrage price at time zero of the option and the hedging portfolio , and , we begin with the last two

Subtracting one of these from the other and solving for mula

and substituting this into either equation, we can solve for

'' e$ ! C' e$ ! ' e$ $ ' # n $ n ' e e$ q C' e$ ! l"@&i4"@&b`Wt@0l! suHt1E@0uP"@&y` A '' e$ ! C' e$ ! ' e$ $ ' # n $ n ' c e$ q C' e$ ! lw@0s4"@0s`r0y! suT1Ed@0uP"@&y` A'' c$ ! C' c$ ! ' c$ $ ' # n $ n ' e c$ q C' c$ ! lwy4wy`4i! suTebPwy` A'' c C' c ' c $'# n $ n ' c c C' c l"d5$ ! "d5$ ! W4d5$ ! suHt1Yd$ q w$ !

' e$ @&y!` ' e$ C' $ ! e$ l"'@0s!4"e0b`Wt'@&i! ' # n ' e e $ q C ' e $ !  ' e e$ $suTne$P@0iP"@0b @&bq u A ' e ! C e$ ' e $ ' n $ y'"@0$s4"'@&y!`4@0$i! su#Ht1nE'd@0iP"@0b 0bq u c e$ q C' e$ !  ' c e$ ' e @&$ ! d5$ ! D ' c

, we obtain the delta-hedging for-

and therefore depend on Although we do not indicate it in the notation, the rst coin toss. Thus, there are really two equations implicit in (1.10):

, the outcome of

q x

! C! 4"`T! ! ! ! 'D CD lsEsGWD $ ' # n $ n ' e $ ! C D x ' e $ u suTt1E@0bxiG @0s! A ' D C D $ ' # n $ n ' c$ ! C D yiEsWD u suTed5bxiG d5y! x ' c $

! qu q PC '! C! $'# n $ n q C!  l"4B`Wt! ibTh1tPB`q u

# n  ' e ('@e0$ q u Ed@0$ q u H 4@0$ ! e n' c e e $ q C ' c e$ q A '@0ee 7PW4'dc@e0$7fC  ' e $ ! ' e $ 0uq u 4d@07q u t@0s`

Du

! ! 4C ' D C D $ ' # n $ n ! C D x llEsGWD u suT1rxsG@! ' e$ 0b!

' c$ ' e$ ! ' c$ ! D d5b! @&b` d5y G D u

D CD ElGD u

(1.10)

@0uq u ' e e$ d@&7q u ' c e$ 7q u ' e c$  ' c c dd5$ q u

qu

(1.11)

(1.12)

(1.13)

16 Equation (1.13), gives the value the hedging portfolio should have at time if the stock goes down between times and . We dene this quantity to be the arbitrage value of the option at time if , and we denote it by . We have just shown that (1.14)

The hedger should choose her portfolio so that her wealth if agrees with dened by (1.14). This formula is analgous to formula (1.9), but postponed by one step. The rst two equations implicit in (1.11) lead in a similar way to the formulas

and

, where

is the value of the option at time if

, dened by (1.16)

This is again analgous to formula (1.9), postponed by one step. Finally, we plug the values and into the two equations implicit in (1.10). The solution of these equations for and is the same as the solution of (1.3) and (1.4), and results again in (1.6) and (1.9). denotes the value at The pattern emerging here persists, regardless of the number of periods. If time of a derivative security, and this depends on the rst coin tosses , then at time , after the rst tosses are known, the portfolio to hedge a short position shares of stock, where should hold

and the value at time outcomes

of the derivative security, when the rst , is given by

coin tosses result in the

1.2 Finite Probability Spaces


Let

be a set with nitely many elements. An example to keep in mind is (2.1)

of all possible outcomes of three coin tosses. Let be the set of all subsets of . Some sets in are , , , and itself. How many sets are there in ?

' e @0$ ! u

(1.15)

 ' c td5$ !

(1.17)

(1.18)

e ! AA! $ '"AsElt1"issldX u SdBsEyx1wssiihdQ u 4T tByxe"sisldbEl u n ' cA! AA! $ # n  '! AA! $! ! AA! Eyx11isssh  o eA! AA! $ C ' cA! AA! $ A '@wsElxe1ssisdXPqf'cAs!ElxeA"sisAs!}$QPC  ! A ! ' eA! AA! $ @"sEyx1"isii}Q u fsElxe"siss}7 u t' Ey 1wssiA ! d$ Ey

x1"isiAs A ! u

e e e e e cA c e cA e c cA c c c s w"w@jwddb e e eA c e eA e c eA c c eA e e cA c e cA e c cA c c c 1w@1w@1wd1@Bww"wdBwdu

('ebq u Ed5uq u fH d5s! u c$ n ' c c $ # n x ' c $ c! ' c$ y! u '@w5uPqfd7PC  ' c $ ! e c$ q C ' c c$ q ' e c$ ' c c$ bq u fd7q u 4b`

e ! @0$ ! ' e ' e e$ ((@0uq u Sd@07q u 4T n ' c e$ # n

! AA! Elt1"isssh

x0y! u  ' e$ ' e$ 0y! u

' y e"sisA ! d$ Ey ! A ! `

D D G ' e$  ' e$ 0y! u u @0s! ' c$ y! u

' c$  ' c$ d5y! u td5s!

e  ! g `

CHAPTER 1. Introduction to Probability Theory

17

(i)

Probability measures have the following interpretation. Let be a subset of . Imagine that is the set of all possible outcomes of some random experiment. There is a certain probability, between and , that when that experiment is performed, the outcome will lie in the set . We think of as this probability.

For example,

which is another way of saying that the probability of

As in the above example, it is generally the case that we specify a probability measure on only some of the subsets of and then use property (ii) of Denition 1.1 to determine for the remaining sets . In the above example, we specied the probability measure only for the sets containing a single element, and then used Denition 1.1(ii) in the form (2.2) (see Problem 1.4(ii)) to determine for all the other sets in .

wG

(i)

Denition 1.2 Let be a nonempty set. A -algebra is a collection following three properties:

c ! Y E E Y p E  e e c A c e c A e c c A c c c A  q p Un p q In Vw"w@j"dddbp

of subsets of

' $ 5

b Qlp  ' $ 5

For

, we dene

on the rst toss is .

A b qq A bq q A 7q

~sp  e e e q 7! @w@sp  e c e  e e c 7! @wbp  e c c u @wdbp q !

Example 1.2 Suppose a coin has probability in (2.1), dene

for

and

for

. For the individual elements of

(2.2)

with the

! P ! fB ' $  lu5 l g

(ii) If

is a sequence of disjoint sets in

, then

A i)6

A ~@sp  c e e q q !  c c e A 7! ! @sp q A 7  c e c q q ! @@wbp  c c c A u! @ddbp

Denition 1.1 A probability measure properties:

is a function mapping

into

with the following

ssiAsBs q A!  $ t' 

' $ 5

18

Here are some important -algebras of subsets of the set

in Example 1.2:

To simplify notation a bit, let us dene

and let us dene

on the rst two tosses

so that

We interpret -algebras as a record of information. Suppose the coin is tossed three times, and you are not told the outcome, but you are told, for every set in whether or not the outcome is in that set. For example, you would be told that the outcome is not in and is in . Moreover, you might be told that the outcome is not in but is in . In effect, you have been told that the rst toss was a , and nothing more. The -algebra is said to contain the information of the rst toss, which is usually called the information up to time . Similarly, contains the information of

Q S BA S BA A A S A S BA Y w BA Y w BA BA A A A A A xYSBrBBA )b

A Q

on the rst two tosses

AQ A Q

on the rst two tosses

A Q

! 

A A A  ! Q4@BA )b

so that

A e e eA c e eA e c eA c c eA e e cA c e cA e c cA c c c A Q1w@lyQww@sQw"w@jbXjd"wdbyA t

A 11www@sQw"w@jjd"wdbyA t e e eA c e eA e c eA c c eA e e cA c e cA e c cA c c c A

and all sets which can be built by taking unions of these The set of all subsets of

on the rst two tosses

A Q

on the rst toss

A X

on the rst toss

e  e e e A c e e A e c e A c c e x s"w1@w@wd@iq c  e e c A c e c A e c c A c c c x b|jw"wddb

! @ P

e e  e e eA c e e s"w@ix  S c e  e c eA c c e sw@"wd@sx  S e c  e e cA c e c wbwBw@wbx  c c  e c c A c c c x dbjdBwddb

(iii) If

is a sequence of sets in , then

! 

w@

A  t A

issAuBAiq@i A!  g  q @  D  !   q @ 

(ii) If

, then its complement

, is also in .

CHAPTER 1. Introduction to Probability Theory

19

contains full the rst two tosses, which is the information up to time . The -algebra information about the outcome of all three tosses. The so-called trivial -algebra contains no information. Knowing whether the outcome of the three tosses is in (it is not) and whether it is in (it is) tells you nothing about

Example 1.3 Let be given by (2.1) and consider the binomial asset pricing Example 1.1, where , and . Then , , and are all random variables. For example, . The random variable is really not random, since for all . Nonetheless, it is a function mapping into , and thus technically a random variable, albeit a degenerate one. A random variable maps into , and we can look at the preimage under the random variable of sets in . Consider, for example, the random variable of Example 1.1. We have

The complete list of subsets of

and sets which can be built by taking unions of these. This collection of sets is a -algebra, called the -algebra generated by the random variable , and is denoted by . The information content of this -algebra is exactly the information learned by observing . More specically, suppose the coin is tossed three times and you do not know the outcome , but someone is willing , whether is in the set. You might be told, for example, that is to tell you, for each set in not in , is in , and is not in . Then you know that in the rst two tosses, there was a head and a tail, and you know nothing more. This information is the same you would have gotten by being told that the value of is . Note that dened earlier contains all the sets which are in , and even more. This means that the information in the rst two tosses is greater than the information in . In particular, if you see the rst two tosses, you can distinguish from , but you cannot make this distinction from knowing the value of alone.

A A A A rYSjBA )  p qC Y @X7HPHGf Hlb7BSf duPC  y  pA ' $ q

we can get as preimages of sets in

q PC 'q C$ BP

qC

'q C$ "P5

q PC

q PC

' $q dbfC

pA i7Sf

Let us consider the interval

. The preimage under

of this interval is dened to be

is:

f  ' $D }EC

 ' e e e$ q C  ' c e e$ q 7t@0bftd@0uPC ASf4'@ewc@0$7qPCP'dcc07P4@w57Prd@w5uPC e e$ q C  ' e e c$ q C  ' c e c$ q A  ' e c c$ q C  ' c c c$ q )s4@wd5uP4dd5uPC

RC

q PC

Denition 1.4 Let be a nonempty nite set and let random variable is a function mapping into .

be the -algebra of all subsets of

AAq A! AD w"sisi@"B"7

Denition 1.3 Let be a nonempty nite set. A ltration is a sequence of -algebras such that each -algebra in the sequence contains all the sets contained by the previous -algebra.

 @

DEC q ! D PC 4C EC

S j 'q C$ "P5h

q PC

 D F  ' e c c$ q svEC t@wd57PC  pq  f  D F WEC !q

q @

. A

20 Denition 1.5 Let be a nonemtpy nite set and let be the -algebra of all subsets of . Let be a random variable on . The -algebra generated by is dened to be the collection of all sets of the form , where is a subset of . Let be a sub- -algebra of . We say that is -measurable if every set in is also in . Note: We normally write simply rather than .

Denition 1.6 Let be a nonempty, nite set, let be the -algebra of all subsets of , let be , and let be a random variable on . Given any set , we a probabilty measure on dene the induced measure of to be

In other words, the induced measure of a set tells us the probability that takes a value in . In the case of above with the probability measure of Example 1.2, some sets in and their induced measures are:

at the number , and a mass of size at the number . A common way to record this information is to give the cumulative distribution function of , dened by

if if if if

By the distribution of a random variable , we mean any of the several ways of characterizing . If is discrete, as in the case of above, we can either tell where the masses are and how large they are, or tell what the cumulative distribution function is. (Later we will consider random variables which have densities, in which case the induced measure of a set is the integral of the density over the set .) Important Note. In order to work through the concept of a risk-neutral measure, we set up the denitions to make a clear distinction between random variables and their distributions. A random variable is a mapping from to , nothing more. It has an existence quite apart from discussion of probabilities. For example, in the discussion above, , regardless of whether the probability for is or .

 ' e e e$ q C  ' c e e$ q G@&bPkd@&7PC

In fact, the induced measure of

places a mass of size

at the number

, a mass of size

(2.3)

q fC fHGs 7 A A  )sjIf A  ' q C $ x  ' Af  STjI A tYHP S$ A  7w ) 6 A l qPC 'Y5$  7 l q q qC !  q ! Vt'rY$ ~uPup SiB)6   q C  A q p A7'  4TB)6  $  ' A A7'  t $  $  ' A 6  ' $  ' )t5 t7$

' wtd$  y

Q p x  @$ '

' $ ' h $ !q

dtd$ ' wA l '  $ c h ' wA $ l l l l

q PC

on , i.e., a way of assigning probabilities The distribution of a random variable is a measure to sets in . It depends on the random variable and the probability measure we use in . If we set the probability of to be , then assigns mass to the number . If we set the probability assigns mass to the number . The distribution of has changed, but of to be , then the random variable has not. It is still dened by

Thus, a random variable can have more than one distribution (a market or objective distribution, and a risk-neutral distribution). In a similar vein, two different random variables can have the same distribution. Suppose in the binomial model of Example 1.1, the probability of and the probability of is . Consider a European call with strike price expiring at time . The payoff of the call at time is the random , which takes the value if or , and takes the value in variable every other case. The probability the payoff is is , and the probability it is zero is . Consider also a European put with strike price expiring at time . The payoff of the put at time is , or . Like the payoff of the call, the payoff of the which takes the value if put is with probability and with probability . The payoffs of the call and the put are different random variables having the same distribution.

(2.4)

Notice that the expected value in (2.4) is dened to be a sum over the sample space . Since nite set, can take only nitely many values, which we label . We can partition the subsets , and then rewrite (2.4) as

Denition 1.7 Let be a nonempty, nite set, let be the -algebra of all subsets of , let a probabilty measure on , and let be a random variable on . The expected value of dened to be

v BP5$ 'q C

is a into

!q

q PC

 ' e e e$ q C  ' c e e$ q 7t@0bftd@0uPC ASf4'@ewc@0$7qPCP'dcc07P4@w57Prd@w5uPC e e$ q C  ' e e c$ q C  ' c e c$ q A  ' e c c$ q C  ' c c c$ q )s4@wd5uP4dd5uPC

AA! )"sislE

e c c  wd

QuYb Y h  p Y 7) lp Y t Y yp "'d$ t ' lp "d$

' Qlp "}$

e e e  g p p c c c ! d p p

s !

CHAPTER 1. Introduction to Probability Theory

21

b) x 

!fB  !fB  !fB  !fB  b) x  Y)~ l"issXlEt! AA ! 

l !

6 c e e  @g ! p

f l

' wA $

c l

v XlH q 5$ 'f C

!q

be is

22 Thus, although the expected value is dened as a sum over the sample space , we can also write it as a sum over .

Denition 1.8 Let be a nonempty, nite set, let be the -algebra of all subsets of , let be a probabilty measure on , and let be a random variable on . The variance of is dened to be the expected value of , i.e., Var

1.3 Lebesgue Measure and the Lebesgue Integral


In this section, we consider the set of real numbers , which is uncountably innite. We dene the Lebesgue measure of intervals in to be their length. This denition and the properties of measure determine the Lebesgue measure of many, but not all, subsets of . The collection of subsets of we consider, and for which Lebesgue measure is dened, is the collection of Borel sets dened below. We use Lebesgue measure to construct the Lebesgue integral, a generalization of the Riemann integral. We need this integral because, unlike the Riemann integral, it can be dened on abstract spaces, such as the space of innite sequences of coin tosses or the space of paths of Brownian motion. This section concerns the Lebesgue integral on the space only; the generalization to other spaces will be given later.

' lsY5$

Var

One again, we can rewrite (2.5) as a sum over , then

rather than over . Indeed, if

takes the values

Q f f f f ~n f Tn s uQ |s XsQ s n f f n uPbp sPbp s XiPbyp ss  q C n y )f  q C l n  q C f 'rY lESj5 sTP@5 ss $ n ' $ f n ' $ e e e ' e e e$ q C n c e e ' c e e$ q C sp w@07fTx@sp "d@07Pn sp "@w@&$ q Irsp "dd@0$ q n e c e ' e c e C n c c e ' c c e C ewbp "@bPIr@wbp "d@w57Pn e c ' e e c$ q C n c e c ' c e c$ q C e c c ' e c c$ q C n c c c ' c c c$ q dbp "@wd57PTrdup "dd7PC

! P ! fB   q ' WS5$ uS p q ' WS5$ 4' $

u) Qlp q ' P'd$ $ x ' $ q' W '`$ "A $

q PC

To make the above set of equations absolutely clear, we consider (2.3). The denition of is

with the distribution given by

qC PG

 qC PG      A B"isiA !

(2.5)

CHAPTER 1. Introduction to Probability Theory

23

, is the smallest -algebra containing all open Denition 1.9 The Borel -algebra, denoted intervals in . The sets in are called Borel sets.

is in , where and are real numbers. Since is By denition, every open interval a -algebra, every union of open intervals is also in . For example, for every real number , the open half-line

is a Borel set, as is

This shows that every closed interval is Borel. In addition, the closed half-lines

and

are Borel. Half-open and half-closed intervals are also Borel, since they can be written as intersections of open half-lines and closed half-lines. For example,

"sisA q A ! b A 

This means that every set containing nitely many real numbers is Borel; if then

Every set which contains only one real number is Borel. Indeed, if

is a real number, then

' $  5

Tu5WPXB1b%wBY ' A $ ' A $  A

! fb BA pQ @b n 

' A $ A $  A lTS5WQw1gSwBY5$

! fb A B4Wp Be$  A

! fb n A IBY tTBY  ' A

XuXb

!f  H

' $  ~

is Borel. Since

is a -algebra, every complement of a Borel set is Borel, so

' A $ ' A Tu5WPXB1$

For real numbers

and , the union

contains

' $  5~

' $  5

Every set which can be written down and just about every set imaginable is in discussion of this fact uses the -algebra properties developed in Problem 1.3.

! Pu 'A lQ4W5$ tXB1$  'A

! fu ' n A TBY5$ rTS$  ' A

' $  5

' $  5

' $  ~

'A lBY5$

' $  5

. The following

24

This means that the set of rational numbers is Borel, as is its complement, the set of irrational numbers. There are, however, sets which are not Borel. We have just seen that any non-Borel set must have uncountably many points. Example 1.4 (The Cantor set.) This example gives a hint of how complicated a Borel set can be. We use it later when we discuss the sample space for an innite sequence of coin tosses.

has two pieces. From each of these pieces, remove the middle half, i.e., remove the open set

The remaining set

has four pieces. Continue this process, so at stage , the set length . The Cantor set

is dened to be the set of points not removed at any stage of this nonterminating process. Note that the length of , the rst set removed, is . The length of , the second set removed, . The length of the next set removed is , and in general, the length of the is -th set removed is . Thus, the total length removed is

and so the Cantor set, the set of points not removed, has zero length. Despite the fact that the Cantor set has no length, there are lots of points in this set. In particular, none of the endpoints of the pieces of the sets is ever removed. Thus, the points

are all in . This is a countably innite set of points. We shall see eventually that the Cantor set has uncountably many points.

! P x p f s s A  s f iA s s s A A )6 q s s s A s s A s  x q

q @

s s issA uf A s A s ssiA q

!  !q f q !

! fB A  p

A s A i 7iA f A f )6 A A A!

fA  )6 !

The remaining set

iA f Pf A f

x ! 

A i)6

Consider the unit interval

, and remove the middle half, i.e., remove the open interval

has

pieces, and each piece has

Aq A!  1sisiQib

In fact, every set containing countably innitely many numbers is Borel; if

, then

!f XuXb  H

!  ! n !

CHAPTER 1. Introduction to Probability Theory

25

Lebesgue measure is dened to be the measure on which assigns the measure of each interval to be its length. Following Williamss book, we denote Lebesgue measure by . A measure has all the properties of a probability measure given in Problem 1.4, except that the total measure of the space is not necessarily (in fact, ), one no longer has the equation

in Problem 1.4(iii), and property (v) in Problem 1.4 needs to be modied to say:

To see that the additional requirment

is needed in (v), consider

We specify that the Lebesgue measure of each interval is its length, and that determines the Lebesgue measure of all other Borel sets. For example, the Lebesgue measure of the Cantor set in Example 1.4 must be zero, because of the length computation given at the end of that example. The Lebesgue measure of a set containing only one point must be zero. In fact, since

for every positive integer , we must have

6  D @)uu

Letting

, we obtain

p  BA G ETbuETG6 n D D

n BA p b

 !  P

Then

, so

, but

 '! $ HB5

tQQ boy  ' $ D 6  ' ! $ D tu P 5QE A' 1issyTBA) lTB)p t@lTB A' A  q A' A  !  '! $ HB5 ! fB ' l)$| yb  ' $  5~

q ! si@H

Aq A! sisi@s

(v) If

is a sequence of sets in

with

and

!P ! fB $ l'i@5  ' $  5~

(ii) If

'' $ A " 5~wG 5$

' $  $ 5dt' 5  ' $ D t 5QE

ssisqBs A A! 6  ' $ t5

(i)

is a sequence of disjoint sets in

, then

'' $ A " 5~wG 5$


, then

A )6

' $  5~

Denition 1.10 Let function mapping

be the -algebra of Borel subsets of into with the following properties:

. A measure on

is a

26 The Lebesgue measure of a set containing countably many points must also be zero. Indeed, if , then

The Lebesgue measure of a set containing uncountably many points can be either zero, positive and nite, or innite. We may not compute the Lebesgue measure of an uncountable set by adding up the Lebesgue measure of its individual members, because there is no way to add up uncountably many numbers. The integral was invented to get around this problem. In order to think about Lebesgue integrals, we must rst consider the functions to be integrated.

Denition 3.4 is purely technical and has nothing to do with keeping track of information. It is difcult to conceive of a function which is not Borel-measurable, and we shall pretend such functions dont exist. Hencefore, function mapping to will mean Borel-measurable function mapping to and subset of will mean Borel subset of .

the set indicated by . We dene the Lebesgue integral of

to be

Let be a nonnegative function dened on dene the Lebesgue integral of to be

 w ' $ ' $ Y5tHSr D x D 7 p)BE7 7 %!f ! )P l'i@$QDy  D @E7iQ x D E7

, possibly taking the value

is simple and

and each

is a real number. We dene the Lebesgue integral of

AX h A Xw

A)6 A  ' $ 7 gtY5QQ

where each

is of the form

if if

A ' $ lY5QQi

%!fB  ' $ tS54

A simple function

from

to

is a linear combination of indicators, i.e., a function of the form

to be

at some points. We

for every

' $ D D l@Qgx  E X r'Y5$r b x

Denition 1.12 An indicator function and . We call

from

to

is a function which takes only the values

' $  ~ ' $  5Wv

' $  5

Denition 1.11 Let

be a function from to . We say that is Borel-measurable if the set is in whenever . In the language of Section 2, we want the -algebra generated by to be contained in .

!f ! f )66 ~uXbuE t5QE   D  ' $ D

' $ Sx qb

 1sisA q A ! b

CHAPTER 1. Introduction to Probability Theory

27 is integrable.

Finally, let be a function dened on , possibly taking the value at some points and the value at other points. We dene the positive and negative parts of to be

respectively, and we dene the Lebesgue integral of

provided the right-hand side is not of the form , since (or equivalently,

Let be a function dened on , possibly taking the value other points. Let be a subset of . We dene

is the indicator function of .

The Lebesgue integral just dened is related to the Riemann integral in one very important way: if the Riemann integral is dened, then the Lebesgue integral agrees with the Riemann integral. The Lebesgue integral has two important advantages over the Riemann integral. The rst is that the Lebesgue integral is dened for more functions, as we show in the following examples. Example 1.5 Let be the set of rational numbers in , and consider set, has Lebesgue measure zero, and so the Lebesgue integral of over

To compute the Riemann integral , we choose partition points and divide the interval into subintervals . In each subinterval there is a rational point , where , and there is also an irrational point , where . We approximate the Riemann integral from above by the upper sum

and we also approximate it from below by the lower sum

s x  ' $ A ! A Aq A! A! A D i)BsE) (isi(1YBBR 1RBu)   !  D  tsiVEUw)V6

A s)6

 x 

! fB !fB A7t' !Ey q 5$X h' !El q s' t $ $ 6h' 5t #$ A! iSBsEyS

6  '! $ )tBEySqS5X@6

Ap h A pw

A s6

6  D )E ! D

!fB ! fB  '! $ ' #$ hBElYWSssi5t

A)6 A x ' 7 gS$

where

x  D A D  v  D H Tn  DE7 D E7 v w AD uE7 D H v x D E@ AQX6By'S5$tQYuY5$ QXlSxu)uY5$ v A { z y x ' A 6 A ' $ { z y x '


to be . If both ), we say that and is integrable. at some points and the value if if

It is possible that this integral is innite. If it is nite, we say that

A i)6 ' $ QwS5t ! D

' $ X"S5t

 )

are nite at

. Being a countable is

28 , the upper sum is always and the lower sum is No matter how ne we take the partition of always . Since these two do not converge to a common value as the partition becomes ner, the Riemann integral is not dened. Example 1.6 Consider the function

It follows that

Now consider the Riemann integral , which for this function is the same as the . When we partition into subintervals, one of these will contain Riemann integral the point , and when we compute the upper approximating sum for , this point will contribute times the length of the subinterval containing it. Thus the upper approximating sum is . On the other hand, the lower approximating sum is , and again the Riemann integral does not exist. The Lebesgue integral has all linearity and comparison properties one would expect of an integral. In particular, for any two functions and and any real constant ,

D 7 ! E n D

 D ! E "

Finally, if

and

' $ ' $ SfY5t

and whenever

for all

, we have

are disjoint sets, then

6 

' $ ! QYS5t E!

A D 7 A7DE7 n D E@ 7

' $ ' $ S5tHY5r

is simple and

6  6D  D )@XbuE  E7

sDES ~DE 7 7  D 7  D' n ESSG5$

A s

' $ XSY5t

E7 D

' $ ! QYS5t E!

)BE7  D

' A TB)6 

for some

, and thus has Lebesgue integral

A) 6  A6  )`

if if

for every

This is not a simple function because simple function cannot take the value function which lies between and is of the form

) 6  A6  )

if if

. Every simple

A )6 x  ' $ A S5t

A )6 x  ' $ A gS54

A s)6

CHAPTER 1. Introduction to Probability Theory

29

There are three convergence theorems satised by the Lebesgue integral. In each of these the situation is that there is a sequence of functions converging pointwise to a limiting function . Pointwise convergence just means that

There are no such theorems for the Riemann integral, because the Riemann integral of the limiting function is too often not dened. Before we state the theorems, we given two examples of pointwise convergence which arise in probability theory.

These converge pointwise to the function

Example 1.8 Consider a sequence of normal densities, each with mean and the -th having variance :

These converge pointwise to the function

We have again for every , so , but . The function is not the Dirac delta; the Lebesgue integral of this function was already seen in Example 1.6 to be zero.

This is the case in Examples 1.7 and 1.8, where

D 7Ei

A  D 7Ei

D yb E7

y b

D E7s

bo y

If

is dened, then Fatous Lemma has the simpler conclusion

D 7i

A pA  A lssB7dBEi

Theorem 3.1 (Fatous Lemma) Let verging pointwise to a function . Then

be a sequence of nonnegative functions con-

6  D d

DE7)i boy ) 6  A )6 A6  ) S5t A x ' $

if if

6 6  D @E@

9 D "87Tby TE7

 D dE7)i

 D @E7Xs

We have

for every , so

, but

& %p x  ' $ 2 ) S5$ 6  D E7)i b`y G w tS5t 6  ' $

for every

Example 1.7 Consider a sequence of normal densities, each with variance mean :

G w
and the -th having

A pA  lsi)B BA
for every

& $ '%p x  ' $ 530)' # Y5i 421 (

' $  ' Y5ttS$ y b

30 . We could modify either Example 1.7 or 1.8 by setting if is even, if is odd. Now if is even, but if is odd. The has two cluster points, and . By denition, the smaller one, , is and the larger one, , is . Fatous Lemma guarantees that even the smaller cluster point will be greater than or equal to the integral of the limiting function. The key assumption in Fatous Lemma is that all the functions take only nonnegative values. Fatous Lemma does not assume much but it is is not very satisfying because it does not conclude that

There are two sets of assumptions which permit this stronger conclusion.

Then

where both sides are allowed to be

Theorem 3.3 (Dominated Convergence Theorem) Let be a sequence of functions, which may take either positive or negative values, converging pointwise to a function . Assume that there is a nonnegative integrable function (i.e., ) such that

Then

and both sides will be nite.

1.4 General Probability Spaces


Denition 1.13 A probability space (i) (ii) (iii)

, a nonempty set, called the sample space, which contains all possible outcomes of some random experiment; , a -algebra of subsets of ; , a probability measure on , i.e., a function which assigns to each set a number , which represents the probability that the outcome of the random experiment lies in the set .

j

A7DEi yb DE7 Q 7 ' $ ' $ Sf S5i

for every

for every

consists of three objects:

p  A isiA)BA7Si D AD 7Ei yb E7 Q 7 G j si'S5QiHY5$7quHSb7IG6 $ ' ' $ !

4 w  D HE7

A pA  A isi)B7 dBSi

Theorem 3.2 (Monotone Convergence Theorem) Let converging pointwise to a function . Assume that

be a sequence of functions

for every

D7 b )By p p  D 7 D 7Ei  D yb E7 ' A  "A $ ' `"A $

DE7 ui"87Ty 9 !fb uDE7 b 7p  6  D E A ' $ s6 5

while but sequence

U D p  igj

CHAPTER 1. Introduction to Probability Theory

31 has the following

In particular, (d) (Continuity from below.) If

We have already seen some examples of nite probability spaces. We repeat these and give some examples of innite probability spaces as well. Example 1.9 Finite coin toss space. Toss a coin times, so that is the set of all sequences of and which have components. We will use this space quite a bit, and so give it a name: . Let be the collection of all subsets of . Suppose the probability of on each toss is , a number between zero and one. Then the probability of is . For each in , we dene

We can dene this way because has only nitely many elements, and so only nitely many terms appear in the sum on the right-hand side of (4.1).

b x  5 Qlp ' $

' $ 5

For each

, we dene

is q H !

(d) (Continuity from above.) If

is a sequence of sets in

with

q ! ss@H

T WPVAHA Ga U  SQAA Ga E lp T R P I H F D x R I F D ' AAq A! $  Qr1wssist1sd x 

 !fB $ l'S5 yb l g isiA q BA ! !fB ' lS5$ yb  Aq A! isisBs

' B $ n ' $  ' $ lCA IP@ A

is a sequence of sets in

' $ r ' $ l5 HA

 @

(d) If

and

are sets in

and

, then

with

(c) (Finite additivity) If

is a positive integer and

l' $ TBisE' ! 5 t' jBis ! 5 n n $  $ AA! "siss ! P ! fB ' $  lu5 l g

are disjoint sets in

isiA q BA !

(b) (Countable additivity) If

6  ' $ t75

(a)

. is a sequence of disjoint sets in , then

Remark 1.1 We recall from Homework Problem 1.4 that a probability measure properties:

, then

, then

, then

(4.1)

32 Example 1.10 Innite coin toss space. Toss a coin repeatedly without stopping, so that is the set of all nonterminating sequences of and . We call this space . This is an uncountably innite space, and we need to exercise some care in the construction of the -algebra we will use here.

, where ranges over the positive In the -algebra , we put every set in every -algebra integers. We also put in every other set which is required to make be a -algebra. For example, the set containing the single sequence

is not in any of the -algebras, because it depends on all the components of the sequence and not just the rst components. However, for each positive integer , the set

We next construct the probability measure on which corresponds to probability for and probability for . Let be given. If there is a positive integer such that , then the description of depends on only the rst tosses, and it is clear how to dene . For example, suppose , where these sets were dened earlier. Then is in . We set and , and then we have

In other words, the probability of a

on the second toss is .

c  ' n $g n q t@ST5 h@  ' $  ' $  ' $  ' $ q @ tS5 5 ' $ 5  Y G  q j

Wv '`"A $

@ 

is also in

on every toss

! fb c u  

c b

is in

and hence in

. Therefore,

on the rst

on the rst

tosses

tosses

on every toss

c  c c c c c b~essdddu

c b

q @

Because sets.

is a -algebra, we must also put into it the sets ,

, and all unions of the four basic

e @e

The set of all sequences which begin with

A c @e

The set of all sequences which begin with

A e c

The set of all sequences which begin with

A c wc

The set of all sequences which begin with

c  q A c  ! ' A A q A! $  x1wt ssib41uxei}gql

c  q A e  ! ' A A q A! $  Wxe@t ssib41uxei}gql

eqx1Awct! ssib41uxei}gql ' A A q A! $ 

e  q A e  ! ' A A q A! $  gxe@t ssib41uxei}gql

For each positive integer , we dene example, contains four basic sets,

to be the -algebra determined by the rst

tosses. For

x

 x  Y  x  S  x 

A i)6

CHAPTER 1. Introduction to Probability Theory

33

for which there is no positive integer such that . Such Let us now consider a set is the case for the set on every toss . To determine the probability of these sets, we write them in terms of sets which are in for positive integers , and then use the properties of probability measures listed in Remark 1.1. For example,

and

According to Remark 1.1(d) (continuity from above),

A similar argument shows that if so that , then every set in which contains and ) has probability zero, and hence very set only one element (nonterminating sequence of which contains countably many elements also has probabiliy zero. We are in a case very similar to Lebesgue measure: every point has measure zero, but sets can have positive measure. Of course, are those which contain uncountably many the only sets which can have positive probabilty in elements.

and we also dene the random variable

is either zero or one, takes values in the interval . Indeed, , and the other values of lie in between. We dene a dyadic rational number to be a number of the form , where and are integers. For example, is a dyadic . For example, rational. Every dyadic rational in (0,1) corresponds to two sequences

The numbers in (0,1) which are not dyadic rationals correspond to a single have a unique binary expansion.

H f t1siwdd@w5$ t1sid5$  ' c c c c c e c  ' e e e e e c c

A is)6

d(isdd5$ ' c c c c E8

Since each

6  ' e e e e t(isd&$

A e  @x A c 

if if

Aq 8A! issiwsP8

In the innite coin toss space, we dene a sequence of random variables

! P d Xp 8  ' ' $ 4d$

6 x ' $ dQE8

c bp

If

, then

on every toss

; otherwise,

on every toss

e c   G6  G6 c bp  @  c up y u @

on every toss

on the rst

tosses

6  @  y b@

c  b

on the rst

tosses

on every toss

on the rst three tosses

AYss cb c b

on the rst toss

on the rst two tosses

by

; these numbers

! fu c b

c u c u

c up 

34 Whenever we place a probability measure on . For example, if we set

First toss is First toss is

First two tosses are

First two tosses are First two tosses are First two tosses are

Continuing this process, we can verify that for any positive integers

we have

In other words, the -measure of all intervals in whose endpoints are dyadic rationals is the same as the Lebesgue measure of these intervals. The only way this can be is for to be Lebesgue measure.

We conclude this example with another look at the Cantor set of Example 3.2. Let be the subset of in which every even-numbered toss is the same as the odd-numbered toss immediately preceding it. For example, is the beginning of a sequence in , but is not. Consider now the set of real numbers

The numbers between can be written as , but the sequence must begin with either or . Therefore, none of these numbers is in . Similarly, the numbers between can be written as , but the sequence must begin with or , so none of these numbers is in . Continuing this process, we see that will not contain any of the numbers which were removed in the construction of the Cantor set in Example 3.2. In other words, . With a bit more work, one can convince onself that in fact , i.e., by requiring consecutive coin tosses to be paired, we are removing exactly those points in which were removed in the Cantor set construction of Example 3.2.

3 ' ed ! A e!d ! $

abI T ` Y

e wc

a Y bI T c

sA6  3 3 3 ew@e c e ceee ' d$ 3 ' !q A ! $ 'd$ Q abI T Yc q '}$ x  3 c c e e e e c d@wc

It is interesing to consider what construct the probability measure

would look like if we take a value of on .

other than

!q

X f @dc  c  e A f @wc A f @We  c A f @ee A p @Wc  A p @e 


and

A i)6 p p Xp A I vX  p A  Xp  H vX G6

q   ' "A $ !

 p ~sA f p  f p A p  p f A p f A  6  p ~sA p p p A  6

h h h h h

A s6 e wc

c @e

on , we have a corresponding induced measure in the construction of this example, then we have

satisfying

when we

CHAPTER 1. Introduction to Probability Theory

35

In addition to tossing a coin, another common random experiment is to pick a number, perhaps using a random number generator. Here are some probability spaces which correspond to different ways of picking a number at random. Example 1.11 Suppose we choose a number from in such a way that we are sure to get either , or . Furthermore, we construct the experiment so that the probability of getting is , the probability of getting is and the probability of getting is . We describe this random experiment by taking to be , to be , and setting up the probability measure so that

The probability measure described in this example is , the measure induced by the stock price , when the initial stock price and the probability of is . This distribution was discussed immediately following Denition 2.8. Example 1.12 Uniform distribution on . Let and let , the collection of all Borel subsets containined in . For each Borel set , we dene to be the Lebesgue measure of the set. Because , this gives us a probability measure. This probability space corresponds to the random experiment of choosing a number from so that every number is equally likely to be chosen. Since there are innitely mean numbers in , this requires that every number have probabilty zero of being chosen. Nonetheless, we can speak of the probability that the number chosen lies in a particular set, and if the set has uncountably many points, then this probability can be positive. I know of no way to design a physical experiment which corresponds to choosing a number at random from so that each number is equally likely to be chosen, just as I know of no way to toss a coin innitely many times. Nonetheless, both Examples 1.10 and 1.12 provide probability spaces which are often useful approximations to reality. Example 1.13 Standard normal distribution. Dene the standard normal density

uDE7o2 @ x ' $ H

Let

and for every Borel set

, dene (4.2)

A 6 bB)5$

This determines for every set . For example, the probability of the interval is , because this interval contains the numbers and , but not the number .

A s6 A i)6

A is)6

~XiQp A f @sp @A f uQp   f  ! c p q y f

& $ f"p

 A i)6 D '5$%DEh'@$ A s6 ' A $  &is)6 ~ A s)6 

x ' $ # S52

' $  ~d

A i)6

f  D gRC

' $  5

' $   5~gg 

' $ 5

A s6

q PC

36

This corresponds to choosing a point at random on the real line, and every single point has probability zero of being chosen, but if a set is given, then the probability the point is in that set is given by (4.2). The construction of the integral in a general probability space follows the same steps as the construction of Lebesgue integral. We repeat this construction below. Denition 1.14 Let i.e., a mapping from If be a probability space, and let , possibly also taking the values

to

be a random variable on this space, .

is an indicator, i.e,

If

is a simple function, i.e,

If

is nonnegative but otherwise general, we dene

is simple and

for every

for every

and

for every

. With this sequence, we can dene

A A pA  A lssB7jSS8

In fact, we can always construct a sequence of simple functions

' }$ d8 ' $

x QS8 yb  X ' $ dRY8 uy tdj8  ' $ ' $8 ' $q8 ' $! 8 sidQYTd7dbPTo6

8 XT8 )B x  Q

where each

is a real number and each

!P fB ! l'u@$ s   Q x v X fB ! A l'}$  ' t}$ y'@5$ xg X 

for some set

, we dene

is a set in

A@w A w

if if

, we dene

such that

& $ fQ p %p # xwBY x A q
g

6  '  ' td$ 4d$

A wBY

' A  "A $

If

in (4.2) is an interval

, then we can write (4.2) as the less mysterious Riemann integral:

CHAPTER 1. Introduction to Probability Theory

37

where then we dene

The expectation of a random variable

The above integral has all the linearity and comparison properties one would expect. In particular, if and are random variables and is a real constant, then

If

for every

, then

In fact, we dont need to have for every in order to reach this conclusion; it is has probability one. When a condition holds with enough if the set of for which probability one, we say it holds almost surely. Finally, if and are disjoint subsets of and is a random variable, then

We restate the Lebesgue integral convergence theorem in this more general context. We acknowledge in these statements that conditions dont need to hold for every ; almost surely is enough. Theorem 4.4 (Fatous Lemma) Let be a sequence of almost surely nonnegative random variables converging almost surely to a random variable . Then

or equivalently,

A X A X8 T X n

! Q ! I Q  X %i n   ' 8 ' d$d$ 'd$d $ 8 '  X8  X H  X QYn $ ' 8

S 9 b "87 y T A X "87Tby T Q 9

Q x  Q x Q 

is dened to be

A pA  A lis)7S

' $ 8 ' dd$

If

is a set in

and

x X I Q v  Q AQX6l'd$ QYuyx'}$ QXld$ )uod$ v A {z  A6A' { z y x ' AH X H Q v A 


is a random variable, we dene

If

is integrable, i.e,

38 Theorem 4.5 (Monotone Convergence Theorem) Let variables converging almost surely to a random variable

Then

or equivalently,

Theorem 4.6 (Dominated Convergence Theorem) Let be a sequence of random variables, converging almost surely to a random variable . Assume that there exists a random variable such that almost surely for every Then

or equivalently,

In Example 1.13, we constructed a probability measure on by integrating the standard normal density. In fact, whenever is a nonnegative function dened on satisfying , we call a density and we can dene an associated probability measure by

We shall often have a situation in which two measure are related by an equation like (4.3). In fact, the market measure and the risk-neutral measures in nancial markets are related this way. We say with respect to , and we write that in (4.3) is the Radon-Nikodym derivative of

The probability measure weights different parts of the real line according to the density . Now suppose is a function on . Denition 1.14 gives us a value for the abstract integral

which is an integral with respec to Lebesgue measure over the real line. We want to show that

AD 2 uE7E

AD 2 uE7o

  X

We can also evaluate

D E

' $ l 5~j

for every

 D 2

p  A lssABA7vdB Y b S y   A QX y u Q is q
almost surely

A pA  A lsi)B BY

be a sequence of random

. Assume that

2 ' A "' $~"G 5$ S yb  A X b y  Q 4 8T

D E X W2  X

X

D E2 5 x ' $

' A ' $ A  l 5"G5$

! G6 7

(4.3)

(4.4)

(4.5)

for in (4.5) and an equation which is suggested by the notation introduced in (4.4) (substitute cancel the ). We include a proof of this because it allows us to illustrate the concept of the standard machine explained in Williamss book in Section 5.12, page 5. The standard machine argument proceeds in four steps.

This is true because it is the denition of

Step 3. Now that we know that (4.5) holds when is a simple function, we consider a general nonnegative function . We can always construct a sequence of nonnegative simple functions such that

and

for every

. We have already proved that

D 2 uE

  X

We let get

and use the Monotone Convergence Theorem on both sides of this equality to

for every

A G

D 2 E7)i  Qi 7 X wj ' $  ' $ Y5Ps uy tY5t ' $ ' $ q ' $ ! isY5QiHS7bIS5bT6

A pA  issB7`BA

where each

is a real number and each

is an indicator function. Then

for every

Step 2. Now that we know that (4.5) holds when is an indicator function, assume that simple function, i.e., a linear combination of indicator functions. In other words,

D 2 bEE  ! )P D2 v Ys  t P ! DE7uY 2  P ! XrY  7 ! )P XYs v  Q t u Y !fB A l'Y5$)Si  ' $ tS5t

D bE2

'5$  ' $ 5

Step 1. Assume that is an indicator function, i.e., that case, (4.5) becomes

for some Borel set

sr bpp

q
' S$ Sx  ' $

CHAPTER 1. Introduction to Probability Theory

39

. In

is a

40 Step 4. In the last step, we consider an integrable function , which can take both positive and negative values. By integrable, we mean that

From Step 3, we have

Subtracting these two equations, we obtain the desired result:

1.5 Independence
In this section, we dene and discuss the notion of independence in a general probability space , although most of the examples we give will be for coin toss space.

1.5.1 Independence of sets

Suppose a random experiment is conducted, and is the outcome. The probability that is . Suppose you are not told , but you are told that . Conditional on this information, the probability that is

The sets and are independent if and only if this conditional probability is the uncondidtional probability , i.e., knowing that does not change the probability you assign to . This discussion is symmetric with respect to and ; if and are independent and you know that , the conditional probability you assign to is still the unconditional probability . Whether two sets are independent depends on the probability measure . For example, suppose we toss a coin twice, with probability for and probability for on each toss. To avoid trivialities, we assume that . Then

' $ A

q ~esp A S Wcsp @ewcbp A q @dbp e e  c c  G6 

v

' $ pA pfT5  5 '  $ x ' $  y

 $  '  $ l'x$ "'5 hwT g j

Denition 1.15 We say that two sets

and

D 2 7EE u DE2 D E72 v 7 7 X v QE

H X A  7 H Q v 7 D u2 AD u2 v

are independent if

 

Q Q v

 g

   c

X

' $ 5

' A  BwA $ ' $ 5

(5.1)

CHAPTER 1. Introduction to Probability Theory

41

These sets are independent if and only if

, which is the case if and only if

If , then , the probability of one head and one tail, is . If you are told that the coin tosses resulted in a head on the rst toss, the probability of , which is now the probability of a on the second toss, is still . Suppose however that . By far the most likely outcome of the two coin tosses is , and . However, if you the probability of one head and one tail is quite small; in fact, are told that the rst toss resulted in , it becomes very likely that the two tosses result in one head and one tail. In fact, conditioned on getting a on the rst toss, the probability of one and one is the probability of a on the second toss, which is .

1.5.2 Independence of -algebras

No matter which set we choose in and which set we choose in the probabilties is the probability of the intersection.

, we will nd that the product of

e cA c c wBwdb

These two -algebras are independent. For example, if we choose the set the set from , then we have

 c c c eA c c e cA c c q @dbp  @wdb rwBwu A q t' S n q s' S n q gW"wup bwBwdbp  $ $  c eA c c e cA c c

e eA e cA c eA c c A Q1wbQd1dbyA )

c eA c c @wdb

Let

be the -albegra determined by the second toss:

contains the sets

from

!  

Example 1.14 Toss a coin twice, and let be given by (5.1). Let determined by the rst toss: contains the sets

 A wf

for every

e eA c eA e cA c c A Q1sQw"dbyA )

Denition 1.16 Let and be sub- -algebras of set in is independent of every set in , i.e,

. We say that

and

are independent if every

be the -algebra

and

e e

!q 

)6 )vpA 6  ' $

!q

S  q p S hfT5  '  $ A p  ' $ ' $ q A w@5 A p  ' $ Y A A  S n q h@5  ' $

7 )6

' $ ' $  '  $ pA "5 hT5

ewbw c   c eA e c  @"u

)6 6 !q

' $ A

e c e cA c c  wBwu

Let set one

and and one . Then

. In words, is the set . We compute

on the rst toss and

is the

!q 

42 Example 1.14 illustrates the general principle that when the probability for a sequence of tosses is dened to be the product of the probabilities for the individual tosses of the sequence, then every set depending on a particular toss will be independent of every set depending on a different toss. We say that the different tosses are independent when we construct probabilities this way. It is also possible to construct probabilities such that the different tosses are not independent, as shown by the following example. Example 1.15 Dene

and for every set , dene to be the sum of the probabilities of the elements in . Then , so is a probability measure. Note that the sets on rst toss and on second toss have probabilities and , so the product of the probabilities is . On the other hand, the intersection of and contains the single element , which has probability . These sets are not independent.

1.5.3 Independence of random variables

Now the pair pair

The set in this last equation is a subset of the plane . In particular, i.e, a set of the form , where and . In this case,

AQGsw   ' 8 $ 8 Gw"A % Hg H  q q H X G"A %p t' $ A ' 8 $  ' 8 "A $ q  A  8 x '  G HvQwUsp x$

takes values in the plane

8Q

Similarly, the measure induced by

A G HQ p @5$ x '

Suppose and on to be

are independent random variables. We dened earlier the measure induced by

is

, and we can dene the measure induced by the

could be a rectangle,

e e cA c e c  D C F  q C wBw@wbiEXSPb

ample, the sets are indepedent sets.

and

ce1Acdc"Awc@ewBAwcdcdcb e e c  F qPC l G 8

Denition 1.17 says that for independent random variables and is independent of every set dened in terms of . In the case of

, every set dened in terms of and just considered, for ex-

of the stock at time In the probability space of three independent coin tosses, the price independent of . This is because depends on only the rst two coin tosses, whereas either or , depending on whether the third coin toss is or .

l G

q PC

Denition 1.17 We say that two random variables and are independent. generate

c c db c eA c c @"wdb ! bq e A c wcBwdcb c@"wcdbp eA c  e cA c c wBwdbp c eA c c  @wdb c b e A c wcdcb !  c b  $ g'  ' $ @   c e  e c A ~sp A @Wsp A p @up @A @dbp  e e  c c e eA c eA e cA c c 1w@1@wBwdb
for the individual elements of to be

and

are independent if the -algebras they

q PC

' 8$ &

l G

' $

is is

CHAPTER 1. Introduction to Probability Theory

43

In other words, for independent random variables and , the joint distribution represented by the measure factors into the product of the marginal distributions represented by the measures and . A joint density for

is a nonnegative function

Not every pair of random variables has a joint density, but if a pair does, then the random variables and have marginal densities dened by

Suppose and have a joint density. Then and are independent variables if and only if the joint density is the product of the marginal densities. This follows from the fact that (5.2) is and , write (5.1) in terms equivalent to independence of and . Take of densities, and differentiate with respect to both and .

which is dened in terms of the random variable . Therefore, this set is in . (In general, we have that every set in is also in , which means that contains at least as much information as . In fact, can contain strictly more information than , which means that will contain all the sets in and others besides; this is the case, for example, if .) In the same way that we just argued that every set in is also in , we can show that every set in is also in . Since every set in is independent of every set in , we conclude that every set in is independent of every set in .

 ' h q $

' 8$ 0

' d$ A

' $

' $ x

' $

' $ ' $ A A '' Qt"d$ riflhdv l $ e  ' $

' d$ @Ah

' $ ' $ A

' 8$  04d

P ROOF : Let us denote and is of the form a typical set in

. We must consider sets in

Theorem 5.7 Suppose and are independent random variables. Let to . Then and are also independent random variables.

 1g A $ 

 A $  X1gH

' fg $

These have the properties

8  G Iv`A  Y'  $ ! A$  '  A A' G HPXSY5$  5$ ' SY'  $ '  $ SS)P5$ A A ' A ' 8 wA $  @XS'  f5$ ! hw5$ A  '  '  P5$ A 8 '  ' lpA$ 5$   8 h Gwvsp uc p  Gs  iw  8
such that

'p5$ ' $ A ' 8$ & ' $ A 8 ' 8$ 04

"A $ ' 8

 ' tY5$

' $ A

' f $

' d$ A

and

and

are independent if and only if

(5.2)

and

be functions from

and

. But

44 be a sequence of random variables. We say that these random Denition 1.18 Let variables are independent if for every sequence of sets and for every positive integer ,

1.5.4 Correlation and independence

provided all the expectations are dened. P ROOF : Let trying to prove becomes

and

be indicator functions. Then the equation we are

The variance of a random variable

is dened to be

Var

The covariance of two random variables Cov

and

is dened to be

Cov Var

Var

For independent random variables, the correlation coefcient is zero. Unfortunately, two random variables can have zero correlation and still not be independent. Consider the following example.

The last claim is easy to see. If almost surely.

and

were independent, so would be

and

, but in fact,

q 8

d  8 d

Example 1.16 Let be a standard normal random variable, let the distribution . Dene standard normal random variable, and are uncorrelated, but

be independent of and have . We show that is also a and are not independent.

According to Theorem 5.8, for independent random variables, the covariance is zero. If both have positive variances, we dene their correlation coefcient

and

which is true because and general functions and .

are independent. Now use the standard machine to get the result for

8 p i W8  i' 8 $ Sp &8s' q &$h "A $ g x ' 8 8 q ' $ x 8 A  8 QwUsp bj p  Gwvsw sw  8

A 8 l'&$rG sE' r @&r"' r $  ' 8$ $

' 8 0$ ' 8 ' $ "A $

8 8 6  uk kp ju Ukp d   d

' $  $ ! '  4

x ' 8 $ j "A P

Theorem 5.8 If two random variables to , then

and

are independent, and if and

A $ sisl' q w q Bl' ! w ! A $

' $ ' q $ ' ! $  ' q ! $ l)5 siPl5 "l5 tQ(isdtr5


are functions from

A A ssisq l!

'S5$ pS5r  ' $

8 q  q

CHAPTER 1. Introduction to Probability Theory


We next check that is standard normal. For , we have

45

Since is standard normal, which shows that is also standard normal. Being standard normal, both Cov and

Where in

We conclude this section with the observation that for independent random variables, the variance , of their sum is the sum of their variances. Indeed, if and are independent and then

This argument extends to any nite number of random variables. If we are given independent random variables , then Var Var Var Var

1.5.5 Independence and conditional expectation.


We now return to property (k) for conditional expectations, presented in the lecture dated October 19, 1995. The property as stated there is taken from Williamss book, page 88; we shall need only the second assertion of the property:

The point of this statement is that if is independent of , then the best estimate of based on the information in is , the same as the best estimate of based on no information.

SG 

(k) If a random variable

' 8 wA $ q 6  6  d l l q d q 8 twA $    ' 8 8 8  p ~  ip  8  p @  p  Q  Qp p  p p  n u p b  Qp I|kp b  p  d n  d  n    ud p  Qp Hud  Usp Tnud  Usp   Usp  ud 8  8 8 8 w 
and and and and , and we have have expected value zero. Therefore, does the measure put its mass, i.e., what is the distribution of ? Var Var is independent of a -algebra

' 8 l0$ n S' p n ' G80$ n 8 )p W8 & q HS' $ $p i q '8p W80nP'G8p W08$s' W STn q ' W ' 8 i q 8p W n

, then

' lX $

n n ' BsiPq $

n' "! $

Q nBsi4q P! $ ' n n "issyq l! AA A

$ $ $ g

    x ' d A$

Var

Var

Var

8 In gd 

i q @ WIA5y ' d d$g

(5.3)

46

The left-hand side of this equation is

, and the right hand side is

The partial averaging equation holds because and are independent. The partial averaging equation for general independent of follows by the standard machine.

1.5.6 Law of Large Numbers


There are two fundamental theorems about sequences of independent random variables. Here is the rst one. be a sequence of independent, identically Theorem 5.9 (Law of Large Numbers) Let distributed random variables, each with expected value and variance . Dene the sequence of averages

We are not going to give the proof of this theorem, but here is an argument which makes it plausible. We will use this argument later when developing stochastic calculus. The argument proceeds in two for every . We next check that Var as . In steps. We rst check that other words, the random variables are increasingly tightly distributed around as . For the rst step, we simply compute

times

For the second step, we rst recall from (5.3) that the variance of the sum of independent random variables is the sum of their variances. Therefore, Var Var

As

, we have Var

j 6 q ' &$ 6 8

 8 o p n  n n m  n n t Ssin T d i IBis4q TP! GSp n

! f  q  q q

6 '8 t)S&$ !f  '8 t)S0$

S8   8

Then

converges to

almost surely as

If is an indicator of some set , which by assumption must be independent of averaging equation we must check is

j "lisA)A7jA p n Bisn q n ! x  8 A A sislq s!

 l'fT5  X %i Q ! $  ! l  ' $ ' $ A "@ X  ' $ ! XSx

f
, then the partial

To show this equality, we observe rst that also check the partial averaging property

Q  X 

is

-measurable, since it is not random. We must

for every

CHAPTER 1. Introduction to Probability Theory

47

1.5.7 Central Limit Theorem


The Law of Large Numbers is a bit boring because the limit is nonrandom. This is because the denominator in the denition of is so large that the variance of converges to zero. If we want to prevent this, we should divide by rather than . In particular, if we again have a sequence of independent, identically distributed random variables, each with expected value and variance , but now we set

As , the distributions of all the random variables have the same degree of tightness, as measured by their variance, around their expected value . The Central Limit Theorem asserts that as , the distribution of approaches that of a normal random variable with mean (expected value) zero and variance . In other words, for every set ,

& $ PX ) %p # ~ kp yb  d s q e Hr |d q 6 td ! P ! fB  ' d t%|A$ q   d # q

then each

has expected value zero and Var

Var

A Pd lsiPPW q PP ! $ |d x # $ n' ' $ n n '

Y8

# S8

48

Chapter 2

Conditional Expectation
Please see Hulls book (Section 9.6.)

2.1 A Binomial Model for Stock Price Dynamics


Stock prices are assumed to follow this simple binomial model: The initial stock price during the period under study is denoted . At each time step, the stock price either goes up by a factor of or down by a factor of . It will be useful to visualize tossing a coin at each time step, and say that

Note that we are not specifying the probability of heads here. Consider a sequence of 3 tosses of the coin (See Fig. 2.1) The collection of all possible outcomes (i.e. sequences of tosses of length 3) is

A typical sequence of will be denoted , and will denote the th element in the sequence . to denote the stock price at time (i.e. after tosses) under the outcome . Note We write depends only on . Thus in the 3-coin-toss example we write for instance, that

Each is a random variable dened on the set . More precisely, let . Then is a -algebra and is a measurable space. Each is an -measurable function , that is, is a function where is the Borel -algebra on I . We will see later that R is in fact 49

PC H

A Aq A! 1Sisis1s ' $ dQPC ' $ dQPC e e eA c e eA e c eA c c eA c c eA e e cA c e cA e c cA c c c Q"w@"w"1d"Bw"wdBwddb

' vV $ u 

PC ' q A ! $ q C t ' A q A ! $ q C t ' $ q l|esdbflxeu|esduPd7PC

A ' ! $ ! C t ' A q A ! $ ! C t ' $ ! l"dyxl1i|esds4db4C

down by a factor of

if it comes out tails ( ).

the stock price moves up by a factor of

if the coin comes out heads ( ), and

D EC

' wA $

h h
! E C

50
= 3 2 S2 (HH) =u S0 = 2 S (H) = uS0 1 1= S0 1 = S (T) = dS0 1 = 2 2 S2 (TT) = d S0 3 = S3 (TTT) = d 3 S 0 = 3 = 2 = 2 S2 (HT) = ud S 0 S2 (TH) = ud S 0 = 3 = 3 = 3 S3 (HHT) = u2 d S0 S3 (HTH) = u2 d S0 S3 (THH) = u2 d S0 3 S3 (HHH) = u S 0

S3 (HTT) = d 2 u S0 S3 (THT) = d 2 u S0 S3 (TTH) = d 2 u S0

Figure 2.1: A three coin period binomial model. measurable under a sub- -algebra of . Recall that the Borel -algebra is the -algebra generated by the open intervals of I . In this course we will always deal with subsets of I that belong to . R R For any random variable dened on a sample space and any , we will use the notation:

Assumption 2.1

2.2 Information
Denition 2.1 (Sets determined by the rst tosses.) We say that a set is determined by the rst coin tosses if, knowing only the outcome of the rst tosses, we can decide whether the outcome of all tosses is in . In general we denote the collection of sets determined by the rst tosses by . It is easy to check that is a -algebra.

z {y

Example 2.1 In the 3 coin-toss example, the collection

of sets determined by the rst toss consists of:

AA pA  BXsis)7G

PC

Note that the random variable

is

-measurable, for each

The sets we dene

etc, are dened similarly. Similarly for any subset

of

q 

 ' XGwx}$  HlxGw t  A X  ' Q  x}$  Hlw  t 

6 t t F  yQ  r lQ   A A

CHAPTER 2. Conditional Expectation


1. 2. 3. 4.

51
,

The collection 1. 2. 3.

4. , 5. The complements of the above sets,

6. Any union of the above sets (including the complements), 7. and .

Denition 2.2 (Information carried by a random variable.) Let be a random variable . is determined by the random variable if, knowing only the value We say that a set of the random variable, we can decide whether or not . Another way of saying this is that for every , either or . The collection of susbets of determined by is a -algebra, which we call the -algebra generated by , and denote by .

these sets are called the atoms of the -algebra In general, if is a random variable

, then

1. 2. 3.

4. Complements of the above sets, 5. Any union of the above sets, 6. , 7. .

h h Q| } | } h hQ'VVeQVh V| h W' 5h'WG Q| } }

Example 2.2 (Sets determined by

) The -algebra generated by

Qj A$ E '  ! ' $ h ' $ '' H "|}$

 $ r' h H

! $ E

If the random variable tion of sets

takes nitely many different values, then

is generated by the collec-

is given by

consists of the following sets:

'd$ T

' t $

' t $

 ! rwP'  $

~ 'VVeQVhV| } 'QQ~ | } 'VSQG~ Q| } } 'WG~ Q| y 'VVQVVQQ~ V| ~ } 'VSQWVSQ|


, , . , , ,

of sets determined by the rst two tosses consists of:

q'  $ !E q r

52

2.3 Conditional Expectation


In order to talk about conditional expectation, we need to introduce a probability measure on our coin-toss sample space . Let us dene

is the probability of ,

the coin tosses are independent, so that, e.g., , .

Denition 2.3 (Expectation.)

If

then

and

We can think of

Let us estimate , given . Denote the estimate by . From elementary probability, is a random variable whose value at is dened by

If or , then even without knowing , we know that

. If we know that . We dene

If or , then even without knowing , we know that

. If we know that . We dene

D EC q

D EC q

C ' c e e$ ' q ! C$  ' e e e$ ' q ! C$ 7DEX td@0llPC 4 t@&sBfC x5j DC  ' $! EX tdbxC ' q }$bPC D  ' $q RC q gduPC c e e  H bEYF4'@ewccsBfC x5j Pddd5sPC 45j DC  $ 'q ! C$  ' c c c$ 'q ! C$ DRCStdbxC F  ' $! Ft'duPC  $q D F  ' $q RC q td7fC e c c  w ' q ! C$ lPC 45j

cddq c c  q PC

If the value of

is known, then the value of

e e e 

' q ! C$ lfC 45j h ' C d$ q  

where

. Properties of

should depend on , i.e., it is a random variable. should also be known. In particular, , then

'q ! C$ BPC 4

2.3.1 An example

' 5j $ b l'd$ "'|}$   $ Q d' 5j w 6 t '  w d$

as a partial average of

A q t@w  ' e c c$

' $ ' l} "d$ t

' q C ! 5j C$ A l'  qPC 45j d8 ! C$  ' $ 8

is the probability of

q ' $ t ' $ r d u 5 h

q PC

! 4C

' egt Ih $ ' A 6$ si)5q h 'q ! C$ PC 45j

h h

etc.

if if

over the set .

, then

$ $

A '' $ q C$  ' $ 'q ! C$ lwdsP5rIh|diBPC 4 ! X"4C  X1lPC x5j  ' q ! C$ ' $   ' $ ' q ! C $ D C' n F bEwQTp5$ !q  QB54G dsPC 45j ! C   D C' n F E"XT$ Y DC n DC EQ q S tEX q ESF q S tEYF q n DC n DC XB4C !

CHAPTER 2. Conditional Expectation

If

, then or

. If we know . We then take a weighted

 ' $q d7fC

S  q Y n q n q Y n q 5  ' $ p C CF C D X ! C D S ! C D XSF DCF  ' $q EXS}7PC e c eA c c eA e e cA c e c  w"jd@1wBw@wbvq

, then we do not know whether average:

For

where

XB4C X"PC 45j !  'q ! C$ 'q C$ P5j h ' q 5 ' q C ! 5j h C$ C$ 'q ! C$ PC 45j A ' $  '  q ! C$ lY5f tSPC 45j qPC 'q ! C$ PC 45j D  EC  DC EQ C q F  C' n F  ' $ D QY D "XT5$ !q 4Sf D F  EC q  DC ESF
For every set is -measurable. is random only through dependence on , has two fundamental properties: if if if

2.3.2 Denition of Conditional Expectation

where is the function dened above.

In other words,

In conclusion, we can write

Let on

Please see Williams, p.83.

The random variable

Then

Furthermore,

be a probability space, and let be a sub- -algebra of . Let be a random variable . Then is dened to be any random variable that satises:

'P j $

' A  Bj"A ' A  j"A

(a)

is -measurable, we dene . We also write 53

54

Existence. There is always a random variable satisfying the above properties (provided that ), i.e., conditional expectations always exist.

Notation 2.1 For random variables

, it is standard notation to write

Here are some useful ways to think about

A random experiment is performed, i.e., an element of is selected. The value of is partially but not fully revealed to us, and thus we cannot compute the exact value of . . Because this estimate Based on what we know about , we compute an estimate of depends on the partial information we have about , it depends on , i.e., is a function of , although the dependence on is often not shown explicitly. If the -algebra contains nitely many sets, there will be a smallest set in containing , which is the intersection of all sets in containing . The way is partially revealed to us is that we are told it is in , but not told which element of it is. We then dene to be the average (with respect to ) value of over this set . Thus, for all in this set , will be the same.

2.3.3 Further discussion of Partial Averaging


The partial averaging property is

(  ' $ u E u

Lemma 3.10 If

is any -measurable random variable, then provided

H E $ ' j u

Note that

is a -measurable random variable. In fact the following holds:

 ' $ 1 E j

We can rewrite this as

'd$e8 'de8 $

' |d$

7@'|}$ 3 8'd$j8 8 lp

Uniqueness. There can be more than one random variable is another one, then almost surely, i.e.,

' |}$

A X  ' $  Q"E j

Q   X8 ' $ E j l"'&$P $ ' $ ' 8 j t 8 j 8


satisfying the above properties, but if

8 "A

8  3 g8

(b) For every set

, we have the partial averaging property

' $ |d"8

H

3 8

(3.1)

(3.2)

, (3.3)

CHAPTER 2. Conditional Expectation

55

Proof: To see this, rst use (3.2) and linearity of expectations to prove (3.3) when is a simple -measurable random variable, i.e., is of the form , where each is in and each is constant. Next consider the case that is a nonnegative -measurable random variable, can be written as the limit of an increasing sequence but is not necessarily simple. Such a of simple random variables ; we write (3.3) for each and then pass to the limit, using the Monotone Convergence Theorem (See Williams), to obtain (3.3) for . Finally, the general measurable random variable can be written as the difference of two nonnegative random-variables , and since (3.3) holds for and it must hold for as well. Williams calls this argument the standard machine (p. 56). Based on this lemma, we can replace the second condition in the denition of a conditional expectation (Section 2.3.2) by:

2.3.4 Properties of Conditional Expectation


Please see Willams p. 88. Proof sketches of some of the properties are provided below.

Proof: Just take

in the partial averaging property to be . is thus an unbiased estimator of the random variable

The conditional expectation of

Proof: The partial averaging property holds trivially when is replaced by . And since is -measurable, satises the requirement (a) of a conditional expectation as well.

(c) (Linearity)

Proof: Take . This set is in since is -measurable. Partial averaging implies . The right-hand side is greater than or equal to zero, and the left-hand side is strictly negative, unless . Therefore, .

6  ' $ `@

'P j $

X k Q P $ ' %H w' ' j $  6 $dlE j  y

6 r ' $ HP j

(d) (Positivity) If

almost surely, then

l4 q $ n ' ' $  ' $ j q IP4 ! j ! hr q q Hn ! ! 5j

If the information content of on is itself.

is sufcient to determine

, then the best estimate of

based

tP j  ' $

6 Hr

6  ' $ 5

(b) If

is -measurable, then

(a)

( u ( j u  ' $

(b) For every -measurable random-variable

, we have (3.4)

! i f  u u

v u

l' $  ' ' $ $ j t"E j

u v u  u

56 (h) (Jensens Inequality) If

Recall the usual Jensens Inequality:

is a sub- -algebra of means that contains more information than . If we estimate based on the information in , and then estimate the estimator based on the smaller amount of information in , then we get the same result as if we had estimated directly based on the information in .

((b) with

If of

is independent of and , then nothing is gained by including the information content in the estimation of .

$  ' l' j h j $

In particular, if

is independent of

(k) (Role of Independence) If

is independent of

, then

, then

d  u

l'P $  ' ' A $ $ j twb5P ' A $ $ El' t5

Take -measurable).

X d  d  ' P $ j d  G 8 5$j  Q8 ' 8 8 ' $ d  E j U8 A X  d  X8 8

(a) (b)

is -measurable;

. Then satises (a) (a product of -measurable random variables is also satises property (b), as we can check below:

Proof: Let only if

be a -measurable random variable. A random variable

is

'f Aj d$

When conditioning on , the -measurable random variable

' $ d  ' y P Aj d$

(j) (Taking out what is known) If

is -measurable, then

acts like a constant.

if and

l' $j t'` P j ' $ $

(i) (Tower Property) If

is a sub- -algebra of , then

' $ $ r $ y"' j 5tfh' x '' $ $ r ' lw 5tfP ' tAj $ $ H ' $ e TXw t
is convex and

, then

 

ixC q X ! ' ' q $ ' "! PC5j "X q q w'Q F C n

nF nF $

$ $

 ! ' q 5j C$

' $ q C' n $  ' $ 'q C$ lduP"X F gtds R5j


A similar argument one time step later shows that

 

)lds"` P5j A ' $ '! q C$ ' $ |di' ! q 5j l ' 5 C$ $

X(' ! q 5j C$

A q QlPC
must be constant on and

QPC  X"" P i q '! q C$  '! q C$  X"" P i


. Notice that

2.3.5 Examples from the Binomial Model

'! q C$ B` f5j tBB4k` A A A  !

CHAPTER 2. Conditional Expectation

Recall that

'! q C$ B` P5j

On the other hand,

A' $ ! C' n $  ' $'! q C$ )lds4"X F tdlB P A ' $ ! C' n $  ' $ '! q C$ w)ldyx"Q F gh|diB P5j H)Slds4"X pF $  A' $! C' n D C F' EY"X n pF $  C  ' $ C$ D QYF n D C q F dl' ! q GH)bEXSF n D AD C  ' $ '! q C$ EC q F tdlB` P5j uDECXSF  q i n D S EC q F q  XlPC

This nal expression is Similarly, We can also write Therefore, We compute Now since

'! C$ " E

We leave the verication of this equality as an exercise. We can verify the Tower Property, for instance, from the previous equations we have

Thus in both cases we have

must satisfy the partial averaging property,

. (linearity) . 57

58

2.4 Martingales
The ingredients are: A probability space

A sequence of random variables

. This is called a stochastic process.

Conditions for a martingale: is -measurable. If you know the information in , then you know the value of 1. Each . We say that the process is adapted to the ltration .

Example 2.3 (Example from the binomial model.) For

we already showed that

For , we set , the trivial -algebra. This -algebra contains no information, and any -measurable random variable must be constant (nonrandom). Therefore, by denition, is that constant which satises the averaging property

The right hand side is

, and so we have

In conclusion,

If If If

then then then

is a martingale. is a submartingale. is a supermartingale.

' bG ! v Aj $

A supermartingale tends to go down, i.e. the second condition above is replaced by ; a submartingale tends to go up, i.e. .

y z

2. For each ,

. Martingales tend to go neither up nor down.

q q q ! q D wGis`xG

A sequence of -algebras , with the property that . Such a sequence of -algebras is called a ltration.

i 

k z

kG y U z SU@

G y z G z

r $ f' ! v Aj

WASsisl!TbD A A p")isss!`wAbG A A D ' A  B`"A $


.

z h e y U

S%SU% W y S%SU% W y S%SU% Wy

 ' qtuG ! v Aj $

Gw Gw fiGw

h h h

Chapter 3

Arbitrage Pricing
3.1 Binomial Pricing
Return to the binomial pricing model Please see:

Example 3.1 (Pricing a Call Option) Suppose (interest rate), . (In this and all examples, the interest rate quoted is per unit time, and the stock prices are indexed by the same time periods). We know that

Find the value at time zero of a call option to buy one share of stock at time 1 for $50 (i.e. the strike price is $50). The value of the call at time 1 is

Suppose the option sells for $20 at time 0. Let us construct a portfolio: 1. Sell 3 options for $20 each. Cash outlay is

2. Buy 2 shares of stock for $50 each. Cash outlay is $100. 3. Borrow $40. Cash outlay is

59

 V h V

if if

if if

W S%

C i

C G U

Cox, Ross and Rubinstein, J. Financial Economics, 7(1979), 229263, and Cox and Rubinstein (1985), Options Markets, Prentice-Hall.

60
This portfolio thus requires no initial investment. For this portfolio, the cash outlay at time 1 is: Pay off option Sell stock Pay off debt

The arbitrage pricing theory (APT) value of the option at time 0 is

Assumptions underlying APT:

Unlimited short selling of stock.

Unlimited borrowing.

No transaction costs.

Agent is a small investor, i.e., his/her trading does not move the market.

3.2 General one-step APT


Suppose a derivative security pays off the amount at time 1, where is an -measurable random variable. (This measurability condition is important; this is why it does not make sense to use some stock unrelated to the derivative security in valuing it, at least in the straightforward method described below).

regardless of whether the stock goes up or down.

We want to choose

and

so that

GGib@uVAVbQS UiyQAU@U

Then wealth at time 1 is

iGC

Invest negative).

iu

Buy

shares of stock at time 0. (

in the money market, at risk-free interest rate . (

Sell the security for

at time 0. (

is to be determined later). is also to be determined later) might be

Important Observation: The APT value of the option does not depend on the probabilities of and .

V Q 

ww U%Q h

CHAPTER 3. Arbitrage Pricing

61

The last condition above can be expressed by two equations (which is fortunate since there are two unknowns):

Note that this is where we use the fact that the derivative security value is a function of , is known for a given , is known (and therefore non-random) at that as well. i.e., when Subtracting the second equation above from the rst gives (2.3)

3.3 Risk-Neutral Probability Measure


! % $# "

Let be the set of possible outcomes from by the formula


!

coin tosses. Construct a probability measure


B E 8 7 5 4 2 0 B @ 8 7 5 4 2 0  FA9) 63D  CA9) 631

% $#

G #$

is called the risk-neutral probability measure. We denote by tion 2.4 says

the expectation under

Then and . Since , we have and . Thus, probabilities. We will return to this later. Thus the price of the call at time 0 is given by







@Q

We have already assumed . We now also assume be an arbitrage opportunity). Dene


 

G  @ iA f Q bUyAAuAb b bh iyA U A GibQfAUAAk



    

W   iA   @ V

 

P f c I

( & ')  UU  '  % $#

G $#

   Q  

 

Plug the formula (2.3) for

into (2.1):

(otherwise there would

are like

(2.4)

on

. Equa-

Ak GibQSfAkx3@bfVS k GibQSfUAVbQS yA yAU i bQS


  

(2.1) (2.2)

 



% $#

s W X W V q b R fVSuQ6 Y6 FR kcSr@Q R kcS 6 ed c6fVS W X W V R


Proof: We have Proof: Theorem 3.12 Under is a martingale.

8 ( T  Sk@VSQ R

% #$

W G bfVSC d x  QS W Uh' UQdU


, which need not be 0.

W 6

UU   A

% R fVS V SU ed cR fVS X W V gf ed cR fVS X W V I h`@ @ ed cR fVS X W V b    H ed cR fVS X W V ` W X W V a6 Y6 FR kVS


b

8 ( U  Sk%SQ T R

The portfolio process is

3.3.1 Portfolio Process

Theorem 3.11 Under

62

is the number of shares of stock held between times

, the discounted stock price process

, where

and

Ci

3.3.2 Self-nancing Value of a Portfolio Process

( h 

G #$

% $#

Each


is

Then each

Dene recursively

Start with nonrandom initial wealth

is

-measurable. (No insider trading).

, the discounted self-nancing portfolio process value -measurable. .

is a martingale.

(3.2) (3.1)

CHAPTER 3. Arbitrage Pricing


Therefore,
G $#

63

3.4 Simple European Derivative Securities


Denition 3.1 () A simple European derivative security with expiration time is an -measurable random variable . (Here, is less than or equal to , the number of periods/coin-tosses in the model).
w A w h

Proof: We rst observe that if martingale property


i

is a martingale, i.e., satises the

for each

, then we also have

When , the equation (4.2) follows directly from the martingale property. For we use the tower property to write
& R w

w A

Theorem 4.13 (Corollary to Theorem 3.12) If a simple European security , the APT value at time of is for each

%v  G   

h 6 D6y G #$ ` W Tv  G    i d  hDQ WS wA w R QS G $# Q i ` i

9 `

&

w hD `

%&

G #$

i 

G #$

G #$

h 6 ` w

&

v 

G #$

v  G   

G #$

v 

GU   

v

In this case, for

, we call

the APT value at time

! w

w h 

U  A

Denition 3.2 () A simple European derivative security a constant and a portfolio process given by (3.2) satises process
x y

is said to be hedgeable if there exists such that the self-nancing value

of

. is hedgeable, then (4.1)

"

R QS R fcS S D6 ` W V cR fVS G $# X W ed R QS S t ` #$ R @ G u S D6 ` W #$ V cR @ G r X W e6 S t ` R VS G #$ ` X W V DdW e6 cR VS


(Theorem 3.11)

(requirement (b) of conditional exp.) (taking out what is known) (property (b))

Q w `

U   

UU 

w A

(4.2) ,

64 We can continue by induction to obtain (4.2).

If the simple European security is hedgeable, then there is a portfolio process whose selfnancing value process satises . By denition, is the APT value at time of . Theorem 3.12 says that

is a martingale, and so for each ,

Therefore,

3.5 The Binomial Model is Complete


Can a simple European derivative security always be hedged? It depends on the model. If the answer is yes, the model is said to be complete. If the answer is no, the model is called incomplete.
w A

Starting with initial wealth is the process

Proof: Let and be dened by (5.1) and (5.2). Set and dene the self-nancing value of the portfolio process by the recursive formula 3.2:

We need to show that

We proceed by induction. For , (5.3) holds by denition of some value of , i.e., for each xed , we have

U  UU  Q%i U  UU  Q U  UU  Q i  i v  GU   Di  T Q x W GUk' UfVSrd 6 W R h  U  w R w  U  R w  UU  w A  U   R h  U   w R fVS G $# w w W  UU 6 W  V  U  Q3d   UU  c3d UV  U  Q U 6 W W

, the self-nancing value of the portfolio process

. Assume that (5.3) holds for

` w UV  U  Q6 A

Theorem 5.14 The binomial model is complete. In particular, let tive security, and set
w R 6

` w Wx A w R @ G #$ fVS Sb `wA w R6fVS G #$ 6 w w 6QS G $# 6QS ` R R i VS  U  R @S 


w w R

be a simple European deriva-

w 1

VS G #$ QS  U  Qi

U 

w A

(5.1)

(5.2)

(5.3)

W GAkdi kdi iAk6   bk6iyAkdi W W W W    kdi iAkWdi I W   W W y  H bk6iyAkdi W k6i   Ak6i  W  bQSf kikdi d A W W W 6 A 6  W   W Gb@uAkdx 3dAk6i W kWW6iAkW6 ibfVSrGb@uAkWd'  U@iA Wd W Akd W W i5kdi   iAkdi    Q

Since will be xed for the rest of the proof, we simplify notation by suppressing these symbols. For example, we write the last equation as

G  U  Q b   U  Qkdi   i  V  U  Qk6i   @ Q W W W U  U  QS `Ddi R VS G #$  U  Qi 8 ( a 6QSQ T R


6

i R kcS S A ` w R 6 G #$ w ` W ` w R 9P6 1 w 6VS G #$ G #$

` W X W V R D6i e6 c6VS

G #$

G   UU  QUdi   U  Qk6 W W W W   UU  QUdi  V  U  Qk6

We need to show that

In other words,

We prove the rst equality; the second can be shown similarly. Note rst that

CHAPTER 3. Arbitrage Pricing

is a martingale under

% $#

We compute

. In particular,

65

66

v z o o
w vx

w  u

 C0 { xu } x  u0 { u0 } { } U u0 }  o n ~ o d c | {
{ |

t u t vt s r p o bhq

%fv  G    %@

Recall that is the given simple European derivative security, and the value and portfolio processes are given by:

4.1 Binomial Model Pricing and Hedging

Example 4.1 (Lookback Option) Consider a simple European derivative security with expiration 2, with payoff given by (See Fig. 4.1):

j i V 6 0m kkhlf d n j ig e   i    GU    v 

W 0f S% U  Qk6y  V  U  Q3d U  U  Q W W UU  QUdiy  V  U  Qk6i W W ` w R i  6 h1 w SkVS G $# QS i


w

Notice that

V x Q x o o y w o o

Using these values, we can now compute:

The payoff is thus path dependent. Working backward in time, we have:

The Markov Property

Chapter 4

67

oo a

a e o o vd S o W"

{ { U G U e  G b |e1  { { g z 5 |eA G G { { h v0h G b |h~ G G


V Q z V o Q oo
o

Example 4.2 (European Call) Let with expiration time 2 and payoff function

Note that

hG  { o ~ G o } G {  h o o ~ o o } o e u h U  o { o } x  o { o }  { o cU o } x V o g Q o o U o

Dene etc.

u 6

Working forward in time, we can check that

68

to be the value of the call at time


S =4 0

Figure 4.1: Stock price underlying the lookback option.

S (H) = 8 1

S1 (T) = 2

when

. Then

S2 (TT) = 1

S2 (TH) = 4

S2 (HT) = 4

S2 (HH) = 16

, and consider a European call

CHAPTER 4. The Markov Property


In particular,
U  o { o } VG z  o { o } {  o F o } U z o z 8 o U Ue o

69

4.2 Computational Issues


i !

For a model with periods (coin tosses), equations of the form


"

has

elements. For period , we must solve

For example, a three-month option has 66 trading days. If each day is taken to be one period, then and . There are three possible ways to deal with this problem: 1. Simulation. We have, for example, that
(  1 f  vyhPU

and so we could compute by simulation. More specically, we could simulate coin tosses under the risk-neutral probability measure. We could store the . We could repeat this several times and take the average value of as an value of approximation to . 2. Approximate a many-period model by a continuous-time model. Then we can use calculus and partial differential equations. Well get to that.
" ( A

& & & & GF  Gc  wxc  xkc

3. Look for Markov structure. Example 4.2 has this. In period 2, the option in Example 4.2 has three possible values , rather than four possible values If there were 66 periods, then in period 66 there would be 67 possible stock price values (since the nal price depends only on the number of up-ticks of the stock price i.e., heads so far) and hence only 67 possible option values, rather than .
f  hh P

W W  V  U  Qk6i     UU  QUdi    Q U  UU  Q W

hG St GSu u e u e

{G $# ( R kcS

& & & USD  D  %D

( AG $# ( ` A (  U  Q

Let

be the number of shares in the hedging portfolio at time

when

. Then

"

70

4.3 Markov Processes


Technical condition always present: We consider only functions on I and subsets of I which are R R Borel-measurable, i.e., we only consider subsets of I that are in and functions R such that is a function .

4.3.1 Different ways to write the Markov property

(If we x and dene , then the equations in (b) and (c) are the same. However in (b) we have a condition which holds for every function , and in (c) we assume this condition only for functions of the form . A main result in the theory of Laplace transforms is that if the equation holds for every of this special form, then it holds for every , i.e., (c) implies (b).)

~` b G $ `

` 3P

` b `

where .)

. (Since

we dont need to assume that

` 6 W

(d) (Agreement of characteristic functions) For every


v ` d W

, we have

hG $

W 6 A6

bG $

G $

(c) (Agreement of Laplace transforms.) For every

for which

, we have

m $ $

e w

(b) (Agreement of expectations of all functions). For every (Borel-measurable) function for which , we have

W d

W S `hd % $ ` W 6 Q6$ G $ b W6 ` Q6$ G $ $

W v6 ` W

G$

` W U h6 % $

` W Sx 6

S 16 W

(a) (Agreement of distributions). For every

, we have

W d

(The Markov Property). For each on is the same as the distribution of


i

, the distribution of conditioned on .

aQ T

W d f"  G   

ckbY`

G $

G $

G$

T U

hv `

The stochastic process

is adapted to the ltration

, and

conditioned

Denition 4.1 () Let be a probability space. Let be a ltration under be a stochastic process on . This process is said to be Markov if:

f h $ $

8 ( aUQ T

 !

P

W ` 6 G $

xC

 !

8 ( a T

R c

. Let

appears, we could just as Remark 4.1 In every case of the Markov properties where well write for some function . For example, form (a) of the Markov property can be restated as:

where is a function that depends on the set . Conditions (a)-(d) are equivalent. The Markov property as stated in (a)-(d) involves the process at a current time and one future time . Conditions (a)-(d) are also equivalent to conditions involving the process at time and multiple future times. We write these apparently stronger but actually equivalent conditions below. Consequences of the Markov property. Let be a positive integer.
$ 4 d W W 6 W W  U  d 6  U  d $ W

(B) For every function

for which

, we have

Remark. All these Markov properties have analogues for vector-valued processes.

Once again, every expression of the form can also be written as function depends on the random variable represented by in this expression.

Y X 7

7 Y hY W |W

hY V

` dUSfG $

G$

S ` X 7 e

u 4 d  $ W

7 Y hY hY V W |W

G$

(D) For every

we have

mh`

WS

7 Y 7 Y

7 e

7 Y hY hY W |W

` hY)fhY G $

G$

S `

7 Y 7 Y W W hY

u 4 d  $ W

U  6 W

hY

G$

(C) For every

for which

WS

S 4 6  U  6 G $ Sx 4 6  U  d G $ ` W W ` W W h ` 4 6  UU  6 G $ m t W W ` $ $ e 4 W U  6 % $ S 4 6  UU  6 ` W W

` h 4 6 W

W W

U  6 W

(A) For every

, we have

, where the

W6

` 4 6 W

4 6 W

W W U  66

(A) For every


%$

,
%$

` W G d % $

W S ` 4 6

ri

For every

, we have

dU `

CHAPTER 4. The Markov Property


G$

71

%$

72
6$ c

Since both and

Example 4.3 It is intuitively clear that the stock price process in the binomial model is a Markov process. We will formally prove this later. If we want to estimate the distribution of based on the information in , the only relevant piece of information is the value of . For example,

is a function of . Note however that form (b) of the Markov property is stronger then (3.2); the Markov property requires that for any function ,
u

Consider a model with 66 periods and a simple European derivative security whose payoff at time 66 is

w Ya

is a function of

. Equation (3.2) is the case of

are equal to the RHS of (3.1)), they are equal to each other, and this is property (A) with

G Q

Now take conditional expectation on both sides of the above equation, conditioned on use the tower property on the left, to obtain

` W W W S W6 G6 d

6 S6 ` W W W G6 d $ G $ & W 9Yd ` W ` & W F6 cd $ G $G6 d $ G $ W W cd 9 6 F6 ` W ` & W & W $ G $G 6 d $ G $ W W c6 W b `96x Gcd W ` & W & W 6 d $ G $ G $ $ W c6 d `v F6 & W P 6 W $ G $ $ W & W `&W &W W W tcd cd  d 6
(Denition of conditional probability) (Tower property) (Taking out what is known) (Markov property, form (a).) (Remark 4.1) (Markov property, form (b).)

e bu

|e u n {

`&W x hF6

` & W F6

S6 G6 d ` W W W

& W F6

& W F6

 u

G$

eu bh }

W 6

W 6

` & W cd

d W

6 W

 bu

& W cd

bu e

%$

%$

G$

W d

d W

Proof that (b) Consider

(A). (with

in (A)) Assume (b). Then (a) also holds (take

%$

u h

).

, and (3.1)

%$ u

(3.2)

CHAPTER 4. The Markov Property


The value of this security at time 50 is
 9

73

because the stock price process is Markov. (We are using form (B) of the Markov property here). In other words, the -measurable random variable can be written as
e 6d 66

for some function , which we can determine with a bit of work.

4.4 Showing that a process is Markov


The following lemma will be very useful in showing that a process is Markov:

Lemma 4.15 (Independence Lemma) Let and . Let be a sub- -algebra of . Assume

be random variables on a probability space

Note that

and

are not independent (unless . If we take , but


o d

or

). For example, one of the sets in from and from


d

is , then

m d d td { d b hd n n

d d td n

d r o S o o

Then and from , then

are independent. For example, if we take and

from

and

Q C

d o Gm V G Q G

e e

 !

Let

be the -algebra of sets determined by the second toss, i.e.,

contains the four sets

c F

i V Q) Gf

Example 4.4 Consider the two-period binomial model. Recall that by the rst toss, i.e., contains the four sets

is the -algebra of sets determined

We say that a random variable by , is independent of .

is independent of a -algebra

if

, the -algebra generated

Denition 4.2 (Independence) Let be a probability space, and let algebras of . We say that and are independent if for every and

Ya

$ G% P % $ f% $

 Y } i |e { |e } |e { {
i

P

 !

and be sub- , we have

is -measurable.

Then
G$

Remark. In this lemma and the following discussion, capital letters denote random variables and lower case letters denote nonrandom variables.
Example 4.5 (Showing the stock price process is Markov) Consider an -period binomial model. Fix a
uhy e fQbu @ u

be any function and set

. Then

The Independence Lemma asserts that


}

Not only have we shown that the stock price process is Markov, but we have also obtained a formula for as a function of . This is a special case of Remark 4.1.
u

4.5 Application to Exotic Options


Consider an -period binomial model. Dene the running maximum of the stock price to be
"

Examples:

( (  cA

( v

"

Consider a simple European derivative security with payoff at time

4 4 a h

 bu

e bu

 bu

e bu

 6u

e bSu

Thus
}

and

are equal and form (b) of the Markov property is proved.

of

u h

This shows the stock price is Markov. Indeed, if we condition both sides of the above equation on use the tower property on the left and the fact that the right hand side is -measurable, we obtain
u h

and

depends only on the

st toss,

uh u { n A h d  6 } u h } 6u I u e bu H

{ 1

{ ~ u

6u dbu  e

 bu

e bbu

time

and dene

and

. Then

if

and

is independent of . Dene

, so that

bu e

G c  ` VhG $

d

Y z e { u h e bu

d

Let

be a function of two variables, and dene

74

is independent of ;

if

. Since

is -measurable. Let




R (  R (

( ( v



G $#

(  c ` ( (

@ @

( A

( ( ( ( 3  cA A %"  G    i
% $# i  S

` W 6i

is Markov. (Here we are working under Lemma 5.16 The two-dimensional process the risk-neutral measure I , although that does not matter). P

the second equality being a consequence of the fact that . Since the RHS is a function of , we have proved the Markov property (form (b)) for this two-dimensional process.

    ib  G5 
d

 


ib1 6 A   d d bA d AG $#

d  d G5U Ayy  A W 6fh  6A dh  6A W W W 6 d d  A% $#    A% $#  W 6 d W  6fy dh W




8 ( 6Uh  AQ T

Stepping back one step, we can compute


G #$

Continuing with the exotic option of the previous Lemma... Let denote the value of the derivative security at time . Since is a martingale under , we have

Q i R fVS

` W A 6

d

cA$ g  cAi ( ( ( ( W ( ( ( ( g g  cAi


W

At the nal time, we have


i

Gh  A

W 6 A

G $#

The Independence Lemma implies


Dene Proof: Fix . We have
i

where

and

CHAPTER 4. The Markov Property

is independent of

indicates the maximum of two quantities. Let

. Let

(Lookback option);

be a function of two variables. We have (Knock-in Barrier option). , so 75

b W  W y  36kgUhU  %6
and the value of the option at time is . Since this is a simple European option, the hedging portfolio is given by the usual formula, which in this case is
 t

d

y  x i W  W k6   A  A36   f c
R ( 

d

( c

A3
R

( A

6

6

( v



 i 1  A

( v



fV

%
R

( v

The general algorithm is so that This leads us to dene 76

Chapter 5

Stopping Times and American Options


5.1 American Pricing
Let us rst review the European pricing formula in a Markov model. Consider the Binomial model with periods. Let be the payoff of a derivative security. Dene by backward recursion:

Now consider an American option. Again a function is specied. In any period , the holder of the derivative security can exercise and receive payment . Thus, the hedging portfolio should create a wealth process which satises almost surely. , and the wealth process
i

American algorithm.

UA

Then

is the value of the option at time .

W W  b 36   1k6   f Q

77

GA

This is because the value of the derivative security at time is at least value at that time must equal the value of the derivative security.
i

%@"  GU    

GA

b 6 zV 6 W W

 i Y x

 GA 

( i

Then

is the value of the option at time , and the hedging portfolio is given by

W k6   iA36k    Q W W

( cA A (

( i

"


v w

w 'c hc w & { b c w e { c b |e| { { e w V 'c h{ w & b c w e { c 5 { { |e e

{ o

vd S o W" Q

Example 5.1 See Fig. 5.1. Then

Let us now construct the hedging portfolio for this option. Begin with initial wealth as follows:
e

78

Figure 5.1: Stock price and nal value of an American put option with strike price 5.

S =4 0

S (H) = 8 1

S (T) = 2 1

S2 (TT) = 1

S2 (TH) = 4

S2 (HT) = 4

S2 (HH) = 16 v2 (16) = 0

v (4) = 1 2

v (1) = 4 2

. Set

. Compute .

w bhp w sr ! #% w { " s r o t o Qbhp 8 v w G

Uw s r bhp ! % s r # o {| o bhp e sr bhp ! " s r o Qbhp

U $ o #

A v 6 A 6  c G $# ` W W


` W W X W V Ux R kcS6 v6Ak6 e6 cR QS5G $#


When do you consume? If The value process is the smallest process with these properties.

G   
i

(0fVSCUbQfQ6Ac QS W ( W  ) kcSd

W 6

w x

& { 'c0 ~ 0 & { '' A 0  & { '' w 1 e  w { 1  e i 5 { { A V o e w & { 'Q w A 0 o w ~ { U i G b { { |1 Q 8 o o z z e x

CHAPTER 5. Stopping Times and American Options

Using

Here,

5.2 Value of Portfolio Hedging an American Option

We get different answers for

Now let us compute

The value satises

" 

i 

 GA 

or, equivalently,

The discounted value of the portfolio is a supermartingale.

is the amount consumed at time .

results in

(Recall that

! If we had

):

, the value of the European put, we would have

. 79

80 and the holder of the American option does not exercise, then the seller of the option can consume to close the gap. By doing this, he can ensure that for all , where is the value dened by the American algorithm in Section 5.1.

In the previous example,

and

. Therefore,

so there is a gap of size 1. If the owner of the option does not exercise it at time one in the state , then the seller can consume 1 at time 1. Thereafter, he uses the usual hedging portfolio

In the example, we have to exercise whenever its value

Denition 5.1 (Stopping Time) Let tion. A stopping time is a random variable

Example 5.2 Consider the binomial model with , so be the value functions dened for the American put with strike price 5. Dene
D e u u h h u 6Y~bt

The stopping time corresponds to stopping the rst time the value of the option agrees with its intrinsic value. It is an optimal exercise time. We note that

We verify that

is indeed a stopping time:

Example 5.3 (A random time which is not a stopping time) In the same binomial model as in the previous example, dene

o n d S o S% " f C  "  GU    i 3 fhAi Q6Q T 4 ! x TmD5"  GG    Dmuf4 Q T  Q ! e 8 ( TaQ P  ! x A bGA bUAk" b W W U k6kxU36k

. It is optimal for the owner of the American option agrees with its intrinsic value . be a probability space and let be a ltrawith the property that:

if if

 1 b A  3 & 3 ` & & fV V8 & 3 Y G|AU G #$ & & & & 1 b%AkU b|AU  2 bxk% D 8 p o E %u W~bA@ C
o

A B

B) B )

az

A@9 7 8

. Let

CHAPTER 5. Stopping Times and American Options


where . In other words, random variable is given by

81

if if if

5.3 Information up to a Stopping Time

Example 5.4 In the binomial model considered earlier, let

The atoms of

are

Notation 5.1 (Value of Stochastic Process at a Stopping Time) If is a probability space, is a ltration under , is a stochastic process adapted to this ltration, and is a stopping time with respect to the same ltration, then is an -measurable random variable whose value at is given by

 !

DV Qf G yB QfQ 7 Q

but

B B

G` X V R '`R R 8 ( U T

The set

is determined by time , but the set

is not. Indeed,

if if

i.e.,

R "

8 p e ku ku %u Y~bA@ 7

B) B )

7 7 Ux 7 Q 7

The collection of sets determined by

is a -algebra, which we denote by

x T 4 Q ii Y fAi Gx

! m

Denition 5.2 Let that

be a stopping time. We say that a set

B V o yB Q

y B

Ux

We verify that is not a stopping time:

V Q) B
is determined by time provided

H I V

PQ

F G A@ D F 8
t t b

p 

stops when the stock price reaches its minimum value. This

8 ( UUQ T

i h q S eS i h p S
H

I o

c H g

V o
I o

G { o 0 o { H o A Q o I H o o

e o S
o o I

o
I o I o

e S c e  c S
o H o I o H o I

c eS  d
o H u

B )

` D

G$

R `X G` YfG $

R W

 " 

 " 

"

8 ( ax  iQ T

(or Theorem 3.17 (Optional Sampling) Suppose that gale. Let and be bounded stopping times, i.e., there is a nonrandom number

8 Ux  T T

If

Taking expectations, we obtain is a supermartingale, then implies If is a martingale, then

82

almost surely, then

, and in particular, . implies

almost surely.

) is a submartinsuch that

. If

8 GT  T

G$

G$

C %" X f W R R V 4 X %R  f ba R V 4 X R


` G $  G$  G$ 

Example 5.5 In the example 5.4 considered earlier, we dene probability measure, the discounted stock price process

for all . Under the risk-neutral is a martingale. We compute

hi u

In every case we have gotten (see Fig. 5.2)


I

H c
,

B f

G G

8 T U  T

V 4

and for
Q

The atoms of

are

and

. Therefore,

CHAPTER 5. Stopping Times and American Options

83

(16/25) S (HH) = 10.24 2

(4/5) S (H) = 6.40 1 (16/25) S2 (HT) = 2.56 S =4 0 (16/25) S2 (TH) = 2.56 (4/5) S (T) = 1.60 1

(16/25) S2 (TT) = 0.64

Figure 5.2: Illustrating the optional sampling theorem.

84

Chapter 6

Properties of American Derivative Securities


6.1 The properties
Denition 6.1 An American derivative security is a sequence of non-negative random variables such that each is -measurable. The owner of an American derivative security can exercise at any time , and if he does, he receives the payment .
i

(b) The discounted value process

is the smallest supermartingale which satises

almost surely.

is an optimal exercise time. In particular


is an optimal exercise time. (d) The hedging portfolio is given by

85

%"  U   

i 

  U  Q36   U  Q%6 W W   U  d    U  6  U  Q W W

T U

R 9r R fVS

r icDA iQ

G #$

f4

(c) Any stopping time

which satises

 t4

where the maximum is over all stopping times

satisfying

 6

` R R 1r R 6QS G $# Q U i R

  i 9 ui x r 8 ( a R fVSQ T

(a) The value

of the security at time

is

8 ( UsQ T r

almost surely.

 " 

U    GU   

i 

 if`i r  u`

8 ( i 6VSQ T R
Q

 " 

i 

8 ( a T

%i R k@VS E GR w S kr ` R R R VS G $# a ` y R r R y R 6fVS G $# ` W R R 9Yd ` y r y R 6VS G #$ G $#

W R y VPd ` y r GR R QSAG #$ 6 ed cR QS W X W V W di


is a supermartingale. , i.e.,

8 ( T SkcSQ R
i i

G #$ E

` R r R 6QS R xr
" 

w GR @ 4

8 ( UxvQ T r

W ` 6 G $# f V V xr Qi

. Then the owner of the derivative security (e) Suppose for some and , we have should exercise it. If he does not, then the seller of the security can immediately consume

Let Dene

6.2 Proofs of the Properties

86

and still maintain the hedge.

be a sequence of non-negative random variables such that each is -measurable. to be the set of all stopping times satisfying almost surely. Dene also

Lemma 2.18

r  ui

Proof: Take

f4

Proof: Let

x '4

Because

Lemma 2.19 The process

is also in

` W 6i eW6 VcR6@ X

G $#

x '4

and

Lemma 2.20 If

attain the maximum in the denition of

to be the constant .

is a supermartingale, then for every . , we have is another process satisfying a.s.,

a.s.

b   U  d     U  6   @ V W W ` W G  U  QS D6i G $# @ V  U  )V  U  (
We prove the rst equality; the proof of the second is similar. Note rst that

W W G   UU  QUdi   U  Qk6 W   U  6 W    UU  d  UU   UU  G  UU  c T U"  UG   Ui Q i i ii x i ( W GU' UQSVd 6 W W  UU c 36 W  V  U  Q3d   UU  ckdi UV  U  Q W U Q k6i W 8 ( Ui R k@VSQ ( T 6i eW6 VcR fVS G $#i R QS q QS ` W X ` W x D6i G #$ @ Q i

(
Lemma 2.21 Dene

% kVS
 

kVS kVS

k@VS E GR w b 1W ` R R R VS G #$ a w ` R G ~r R R VS G #$ aE R




i
G $#

`R 1W R R Q

x "f4 Y R k@VS 8 ( i 6QSQ T R

We need to show that Proof: Then Set Since Therefore,

Proof: The optional sampling theorem for the supermartingale

CHAPTER 6. Properties of American Derivative Securities

We proceed by induction on . The induction hypothesis is that , i.e., for each xed we have and dene recursively is a supermartingale,

must be non-negative almost surely. Dene

implies

for some 87

88 will be xed for the rest of the proof, we will suppress these symbols. For Since example, the last equation can be written simply as

We compute

6.3 Compound European Derivative Securities


In order to derive the optimal stopping time for an American derivative security, it will be useful to study compound European derivative securities, which are also interesting in their own right. different simple European derivative A compound European derivative security consists of securities (with the same underlying stock) expiring at times ; the security that expires at time has payoff . Thus a compound European derivative security is specied by the process , where each is -measurable, i.e., the process is adapted to the ltration . Hedging a short position (one payment). Here is how we can hedge a short position in the th European derivative security. The value of European derivative security at time is given by
i   " 

and the hedging portfolio for that security is given by

Thus, starting with wealth time we have wealth .


4

, and using the portfolio

, we can ensure that at

Hedging a short position (all payments). Superpose the hedges for the individual payments. In other words, start with wealth . At each time , rst make the payment and then use the portfolio
T U

%  UU 

W GAk6 k6 AUdi   5kdiAkdiA W W W W   W W k6i iAkdi    W W GbfVSC kdi5yA k6 ( U)'iAk@QS W W kW6iAkWd UbQSfAkdx kdA%6i W ( W kU' UQSiAkd Ak6 W W W 5kdi   iAkdi    Q ) ( "  G  


X R 4 V 4  U  X 4 V A V X 4   U  Q X 6V4W   V  U  Q X 6V4W i  X X   U  c 4 dVW   V  U  Q 4 6VW  U  Q X 4 V

U  

GU  

X U(

8 4 T 4

 Q Di

fUUC

( v

 6

"

` 4 ( 4 R fVS G $# Q X 4 V

X & W Pcd

X W V f Y6 V

X 4

8 4 (

G  UU  Q (

UQ

8 U T ( 8 4 T 4 (

( s
Q

CHAPTER 6. Properties of American Derivative Securities


( C

89 , we

and the hedging portfolio is . You can borrow and consume it immediately. This leaves you with wealth . In each period , receive the payment and then use the portfolio . At the nal time , after receiving the last payment , your wealth will reach zero, i.e., you will no longer have a debt.

6.4 Optimal Exercise of American Derivative Security


In this section we derive the optimal exercise time for the owner of an American derivative security. be an American derivative security. Let be the stopping time the owner plans to Let use. (We assume that each is non-negative, so we may assume without loss of generality that the owner stops at expiration time if not before). Using the stopping time , in period the owner will receive the payment In other words, once he chooses a stopping time, the owner has effectively converted the American derivative security into a compound European derivative security, whose value is

The owner of the American derivative security can borrow this amount of money immediately, if he chooses, and invest in the market so as to exaclty pay off his debt as the payments are

Lemma 4.22

is maximized by the stopping time

Proof: Recall the denition

| `

Gupy sGGw vUGw ~ s Gb } sw | Gbt w k w

received. Thus, his optimal behavior is to use a stopping time

which maximizes

pe} am ip l sj sw

f ge

$ ( fVS
4 4 R

Guq0 { vw k w ~ s Gpy am p pk xyp m 0wuvs  ~s p i Gpypam p  ~ s sw } | i

l k dj { Gf 2 f 6G`t'n

$pe p

( s
Q

Suppose you own a compound European derivative security


( X 4 (

. Compute

corresponding to all future payments. At the nal time , after making the nal payment will have exactly zero wealth.
8 4 T 4 ( "

i e

{p x us p Uzk yp m 0wvtrqe

8 4

G #$

8 T R ( akQ "

8 4

f k

a
"

} | sw

m l k a6fi Gf vj

f h

f k f m l afi Gf f Guy j
, the

It follows that also attains the maximum in (4.1), and is therefore an optimal exercise time for the American derivative security.

Gu0 Gu0 { t w k sw ~ s w k w ~ s
# # #

Gu0 { w w k w ~ s Gu0 w w sw Gu py~ s Gu0 w k Gw w Gw Gf 2 f 6G`t n l k dj


a.s.
"

w 2 w k
#

Gq0 sGGw vw k w ~ s bt
attains the maximum in the formula

# # #

w w w { A w k sw Gpy

{ Gu0 Ga k w ~s Gu0 sw ~s Gu0 w ~s } sw |

Let be a stopping time which maximizes , i.e., Because is a supermartingale, we have from the optional sampling theorem and the inequality following:

~ s
#

Therefore,

qA w k w

Gpy

~ s w w
#

w k2 w Gq0

~s
#

But we have dened

We have just shown that if

Taking expectations on both sides, we obtain and so we must have

Y n  n

then

and

90

almost surely. The optional sampling theorem implies a.s. (4.1)

{ ~ s ~ s Gb ~ s
This implies
is linear

W W G b
is linear

Jensens Inequality

Chapter 7

7.1 Jensens Inequality for Conditional Expectations

Lemma 1.23 If

{ 9~ s 9~ s a W j l Gd { ~ s ~ s v ~ s  h s se
is convex and :

Proof: Since

Now let

For instance, if

is convex we can express it as follows (See Fig. 7.1):

lie below . Then,

u ~ s s u ~ ~ s ~ s

91

u9

W%

, then

92

Figure 7.1: Expressing a convex function as a max over linear functions.

Proof:

7.2 Optimal Exercise of an American Call


This follows from Jensens inequality. Corollary 2.25 Given a convex function where . For instance, is the payoff function for an American call. Assume that . Consider the American derivative security with payoff in period . The value of this security is the same as the value of the simple European derivative security with nal payoff , i.e.,

Proof: Because

is convex, for all

we have (see Fig. 7.2):

"

bn

where the LHS is the European value and the RHS is the American value. In particular optimal exercise time.

is an

% Uq0 w w ~ s Gt i w % i

{ f f f ~ s f 6f ~ s a ytYa { {  0t W {  d  6 rC s e
#

Theorem 1.24 If

is a martingale and

is convex then

m a6fi l f Uj

Guy

~ s

m l a6fi f sj
#

is a submartingale.

Therefore,

and this last expression is the value of the European derivative security. Of course, the LHS cannot be strictly less than the RHS above, since stopping at time is always allowed, and we conclude that
# # #

f f f

Gu0 Gq0 { i i 2~ s w w w ~ s Gb Gu0 Gq0 i i 2~ s w w w ~ s Gb Gq0 { i G q0 ~ s Uu0 i ~ s g~ s w w ~ s vw i i Gq0 Gpy { vw i i ~ s w w m a6fi l f f Uuq0 j Gq0 n f f p Gq0 6f ~ s t f p Gq0 6f ~ s f % u Gq0 % Gu0 6f ~ s f f 6f } 6f | ~s % u  6f $f
# # # P # # #

So is a submartingale. Let optional sampling theorem implies

and

CHAPTER 7. Jensens Inequality

( x, g( x))

( x, g(x))

Figure 7.2: Proof of Cor. 2.25

be a stopping time satisfying

"

Therefore, the value of the American derivative security is Taking expectations, we obtain

(x,g(x))

. The 93

94
S2 (HH) = 16

S (H) = 8 1 S2 (HT) = 4 S =4 0 S1 (T) = 2 S2 (TH) = 4

S2 (TT) = 1

Figure 7.3: A three period binomial model.

7.3 Stopped Martingales

Example 7.1 (Stopped Process) Figure 7.3 shows our familiar 3-period binomial example. Dene if if

Then

Theorem 3.26 A stopped martingale (or submartingale, or supermartingale) is still a martingale (or submartingale, or supermartingale respectively).

l vv{ U {{ j

Proof: Let The set

{ w 6f f x f G0wu s w x f G0wu s f 6f~ s x f Gw0u s w x f G0wu s u u f f x 6f G0wvs w x f G0wv s ~ s f w } f | ~ s t Ul Uj gl nUj f n f l Uj n m l afi f sj

be a martingale, and is in , so the set

 qq   Ip p pE ` G0' bE bb bp 2YqA' qp { {{ vs{ bd } | w f


if if if if

be a stopping time. Choose some is also in . We compute

Let be a stochastic process and let stopped process

be a stopping time. We denote by

m afi l w f sj

m l a6fi Gf vj

the

CHAPTER 7. Jensens Inequality

95

96

Chapter 8

Random Walks
8.1 First Passage Time
Toss a coin innitely many times. Then the sample space is the set of all innite sequences of and . Assume the tosses are independent, and on each toss, the probability of is , as is the probability of . Dene

The process Brownian motion. Dene

is a symmetric random walk (see Fig. 8.1) Its analogue in continuous time is

If never gets to 1 (e.g., ), then . The random variable is called the rst passage time to 1. It is the rst time the number of heads exceeds by one the number of tails.

It is shown in a Homework Problem that




( & ( & f (& f 120 f ( & )'%#! f ( & $ "     m l  af vf j af Gf j m l 

8.2

is almost surely nite


and where

97

m { s{v{ ' ' p f p f ' p

p p

if if

{{{ svf n { l f U` n j

ml Wff j

{{ vs{ f

98

Mk k

e e + 2 1
Figure 8.2: Illustrating two functions of


for every xed (See Fig. 8.2 for an illustration of the various functions involved). We want in (2.1), but we have to worry a bit that for some sequences , . to let

Furthermore,

, because we stop this martingale when it reaches 1, so

n 2 n

( &

w f

( &

&

( &

w 6f

'

As

A C #D@ B@ A w f w (& (& {

We consider xed

, so

if if

are martingales. (Take in part (i) of the Homework Problem and take (v).) Since and a stopped martingale is a martingale, we have
7 8

w f

f
2 e e + 1
in part (2.1)

Figure 8.1: The random walk process

&

( &

w f

( &

5 6

 ~ s w f 4~ s

bf

9 2

'

CHAPTER 8. Random Walks

99

In addition,

For all

i.e.,

We know there are paths of the symmetric random walk which never reach level 1. We have just shown that these paths collectively have no probability. (In our innite sample space , in each path individually has zero probability). We therefore do not need the indicator (2.2), and we rewrite that equation as (2.3)

Solution:

( & ( & & X ( & X

&

( &

s X

Let

be given. We want to nd

so that

8.3 The moment generating function for


9 2

n l Uj s

m l af Gf j nj { l U9W s n l Uj s ~ s

( & ( & { ( & ~s w

so we can let

U V

'

Letting

, and using the Bounded Convergence Theorem, we obtain (2.2)

, we have

in (2.2), using the Bounded Convergence Theorem again, to conclude

( & ( & & x Tyu s ( & S w w ( & ( & S { " x Tw0u s ( & ~ s w

w f

Recall Equation (2.1):

{ n 2 n

{(& w
if if
A EC @

A@

w 6f

and

( ( )& (

&
&

( &

( &

w f

w f
(

Q I G F RPHB( &

&

( &

5 6

( &

~s w 6f
( &

A f

` { XX ` X 0 X ` $ X X d X X w n d ~ s w 4~ s
X

X
`

w 2~ s X


w 2~ s X
X (

( & ( & X X `

&
9 2

{
9 2

( & ( & ( & ~s

{
X

X
`

(&

X ` X X X b X bc X 0 0 X X (& X (& {
X

` Y a

Recall Equation (2.3):

8.4 Expectation of

9 2

We want

With

We take the negative square root:

100

, so we must have

and

s X

so Recall that this becomes

We have computed the moment generating function for the rst passage time to 1. related by . Now , so (3.1)

CHAPTER 8. Random Walks


e fX

101 in the equation

Using the Monotone Convergence Theorem, we can let

to obtain

Thus in summary:

8.5 The Strong Markov Property


The random walk process

is a Markov process, i.e.,

random variable depending only on same random variable

In discrete time, this Markov property implies the Strong Markov property: random variable depending only on same random variable

for any almost surely nite stopping time .

8.6 General First Passage Times


Dene

Then is the number of periods between the rst arrival at level 1 and the rst arrival at level 2. The distribution of is the same as the distribution of (see Fig. 8.3), i.e.,

{f {{ f sv{ f f

{ vw {{ sw ss{ w w

n {{ vs{ i h l

v { EX

X X

{ "~ s n l U9W s nj l f U` n j

{ "~ s n ` X w n ~ s X
` X h c

f ' Uj ` n g

w qpw 2~ s X

~ s ~ s m l Wf f j ~ s
r s

~ s n

vn

, and payoff function

f f '
0 g X

s { X 0
X 0 X

X
`

Dw 2~ s X
In general,

X
`

X
`

{ sw 4~ s Bpw 2~ s $ X X r sw qpw tX~ s r


, Take expectations of both sides to get

{ 0 X X ` $ X sr w sw r r Pw Bpw tX~ s sw X r r Pw Pw Bpw tX~ s sw X r r X t~ s X Pw r r sw r Pw Bw p u xyPw r r w Bpw X sw wXv~ s


r P

For

Consider the binomial model, with neutral probabilities are ,

8.7 Example: Perpetual American Put

102

Mk

Figure 8.3: General rst passage times.

, and thus

(strong Markov property)

(taking out what is known)

2 1

not random

. The risk

v EX

CHAPTER 8. Random Walks


W#

103 . Suppose

Rule 0: Stop immediately.

Rule 1: Stop as soon as stock price falls to 2, i.e., at time

Rule 2: Stop as soon as stock price falls to 1, i.e., at time

Because the random walk is symmetric under , has the same distribution under as the stopping time in the previous section. This observation leads to the following computations of value. Value of Rule 1:
W#

Value of Rule 2:

This suggests that the optimal rule is Rule 1, i.e., stop (exercise the put) as soon as the stock price falls to 2, and the value of the put is if . Suppose instead we start with , and stop the rst time the price falls to 2. This requires 2 down steps, so the value of this rule with this initial stock price is

We dene

In general, if for some , and we stop when the stock price falls to 2, then steps will be required and the value of the option is
#

{ { r C w G~ u  0 s u ~s r Cw r Cw g n Ws j { l f U` n

{{ vs{ ' o p { c p h { p { g wfr DCe dC w ~ s o { { c p C w ~ s

{l

{ { p C w ~ s s $p C w | }

j f U` g n

$ w | g n } g
#

}r C w|

where

is a symmetric random walk under the risk-neutral measure, denoted by . Here are some possible exercise rules:

down

104 for some , then the initial price is at or below 2. In this case, we exercise If immediately, and the value of the put is

Proposed exercise rule: Exercise the put whenever the stock price is at or below 2. The value of as we just dened it. Since the put is perpetual, the initial time is no this rule is given by different from any other time. This leads us to make the following:

How do we recognize the value of an American derivative security when we see it? There are three parts to the proof of the conjecture. We must show: (a) (b) (c)

is a supermartingale,

is the smallest process with properties (a) and (b).

Proof: (a). Just check that


p p h p { c p h

for

This is straightforward.

Proof: (b). We must show that

By assumption,
h

for some . We must show that

If

, then

and

{ o o

for

Note: To simplify matters, we shall only consider initial stock prices of the form always of the form , with a possibly different .

{ p h { { p { { p { { p {{ p Y p p h h { g p Y p { p p p h h oh p gf { f h { f % h { f 6f ~ s f

p p

Conjecture 1 The value of the perpetual put at time

{{ vv{ ' o p
is
h

h h

m a6fl f j m l af m f h k f j f h f i h

o p

p u

, so

is

CHAPTER 8. Random Walks


c

105 and

There is a gap of size .


h

There is a gap of size 1. This concludes the proof of (b). Proof: (c). Suppose (a) (b)

is some other process satisfying:

is a supermartingale.

We must show that

Actually, since the put is perpetual, every time

provided we let in (7.2) be any number of the form . With appropriate (but messy) conditioning , the proof we give of (7.2) can be modied to prove (7.1). on

Suppose now that

for some

j {l o f U` l f U` n j o  p { p W h

so if

for some

, then (a) implies

, i.e.,

. Let

For

{ p p vt p p h p p p p

p  h % {  f 9f

is like every other time, so it will sufce to show (7.2)

ml aff j f i f  f m l afi Gf sj

p q

If

, then

and

g  n n # h o v0 { { { Y h ph p
(7.1)

o f

If

, then

106 Then

Because

is a supermartingale

Comment on the proof of (c): If the candidate value process is the actual value of a particular exercise rule, then (c) will be automatically satised. In this case, we constructed so that is and if we exercise the put the rst time the value of the put at time if the stock price at time is ( , or later) that the stock price is 2 or less. In such a situation, we need only verify properties (a) and (b).

8.8 Difference Equation


If we imagine stock prices which can fall at any point in , not just at points of the form for integers , then we can imagine the function , dened for all , which gives the value of the perpetual American put when the stock price is . This function should satisfy the conditions:
d

(c) At each , either (a) or (b) holds with equality. In the example we worked out, we have For

This suggests the formula


h

We then have (see Fig. 8.4): (a) (b)

for every

W { a w u v { p p he o s p t p { g p e o h

a W w h W h p a h i

(b)

a i

W r D q ah

(a)

For

except for

{ w h

{ w w ~ s p { g p h 6 f

w ~ s w w ~ s 9

m f f al Y f j o

W a h

CHAPTER 8. Random Walks

107

v(x) 5
(3,2)

Figure 8.4: Graph of

Check of condition (c):

If or , then both (a) and (b) are strict. This is an artifact of the discreteness of the binomial model. This artifact will disappear in the continuous model, in which an analogue of (a) or (b) holds with equality at every point.

8.9 Distribution of First Passage Times

we recall that

We will use this moment generating function to obtain the distribution of . We rst obtain the Taylor series expasion of as follows:

Let

be a symetric random walk under a probability measure

, with

s y

l f U` n j

s Ya

w 2~ s X

X w 2~ s

ml afGf j

If

, then (b) holds with equality:

If

, then (a) holds with equality.

W%

x x x

. Dening

{
0

m { l o G9W s p X p w 2~ s nj X s o o m p $ X o p X X X z w 2~ s X
X

{p

v o o m p $ p o  8s o  o m p p m o p $  o am p p } p | z ` W%

W
z

 ms o  8 o p  8s u } o p} } { E} p } ~ } {sv{ { {{ sv{ o o E} } ~ | } {p { W 0 o sv{ } W } p z p o E} p s{ v{ } ~ } p | z { f C wr De p {{ { W 0 sv{ W | z p W z W 0 z W 0 W z z ` a z z z


108

But also, So we have

The Taylor series expansion of

is given by

CHAPTER 8. Random Walks

109

Figure 8.5: Reection principle.

Therefore,

8.10 The Reection Principle

To count how many paths reach level 1 by time , count all those for which double count all those for which . (See Figures 8.5, 8.6.)

{{ vv{ ' o

s o $ o l o U9W s nj o p l U9W s nj

Figure 8.6: Example with

and

g p

{ l q l p j9W s p 9W s l l j l ag p 9W s l j

s { 0 o o $ o p  8s  o o  8 o  ss 8s  p  o o o o 8 o o o ss s 8 s  p o o o o 8 o o o p 8 ms  ms   o o ms o o m o p o p j j l q p 9W s l p 9 W s p 9W s j j 9W s l l r p l o U9W s l o G9W s l o U9W s nj nj nj o p 9W s j p 9W s j r p 9W s l o U9W s j nj


,

For 110

In other words,

Chapter 9

Pricing in terms of Market Probabilities: The Radon-Nikodym Theorem.


9.1 Radon-Nikodym Theorem
Theorem 1.27 (Radon-Nikodym) Let I and P be two probability measures on a space . satisfying , we also have . Then we say that Assume that for every is absolutely continuous with respect to I . Under this assumption, there is a nonegative random P variable such that (1.1)

Remark 9.1 Equation (1.1) implies the apparently stronger condition

Remark 9.2 If is absolutely continuous with respect to I , and I is absolutely continuous with P P respect to , we say that I and are equivalent. I and are equivalent if and only if P P

(Let

and

be related by the equation

to see that (1.2) and (1.3) are the same.)

111

If I and P are equivalent and is the Radon-Nikodym derivative of Radon-Nikodym derivative of I w.r.t. , i.e., P

w.r.t. I , then P

{  i ~ s 9~ s {

~ s

exactly when

Ws { d i W s

W#

~s

W#

W#

for every random variable


W#

for which

~ s ~ s ~ s

W#

W#

W#

and

is called the Radon-Nikodym derivative of

with respect to I . P

is the (1.2) (1.3)

$z

W#

t d i W s

Tg

W#

W#

W#

112
Example 9.1 (Radon-Nikodym Theorem) Let of length 2. Let correspond to probability for

and

for . Then

, so

9.2 Radon-Nikodym Martingales


Let be the set of all sequences of coin tosses. Let I be the market probability measure and let P be the risk-neutral probability measure. Assume

Dene the I -martingale P

Proof:

{ W s f

Ws

or equivalently,

f d

s f t~ s

s~ s

Note that Lemma 2.28 implies that if


#

is

-measurable, then for any

{ f ~ s { f t~ s)~ s f ~ s~ s ~ s

~s

Lemma 2.28 If

is

We can check that

is indeed a martingale:

-measurable, then

f ~ s ~ s {f f t~ s f f t~ s~ s f 6f t~ s { {vv{ f ~ s f { { WW ss

W#

W#

so that I and P

are equivalent. The Radon-Nikodym derivative of

with respect to I is P

q vb v`
9
#

 i W s W s

and

for , and let

, the set of coin toss sequences correspond to probability for

q b q`0bbv

b p pA

W#

CHAPTER 9. Pricing in terms of Market Probabilities


Z2 (HH) = 9/4 1/3 Z (H) = 3/2 1 1/3 Z =1 0 2/3 Z2 (HT) = 9/8 1/3 Z (T) = 3/4 1 2/3 Z2 (TT) = 9/16 Z2 (TH) = 9/8

113

2/3

(Lemma 2.28)
W

(Partial averaging)

(Lemma 2.28)

9.3 The State Price Density Process


In order to express the value of a derivative security in terms of the market probabilities, it will be useful to introduce the following state price density process:

Example 9.2 (Radon-Nikodym Theorem, continued) We show in Fig. 9.1 the values of the martingale We always have , since

p t

Proof: Note rst that

{ vs{ f f {{

{ p f~ s pp p ~ s o
d

Ws W s  f d d s p f ~ s T W s p f ~ s pp

Uq0

f p p ~ s p

Lemma 2.29 If

is

-measurable and

Figure 9.1: Showing the are for I , not . P


W#

values in the 2-period binomial model example. The probabilities shown

, then

is

-measurable. So for any

, we have

o i pk6p p p am ip l p p j

{ p w k w ~ s e  GiG w p Gq0 Gq0 ~ s pp e  GG w p p w k w w Gu0 Gq0 p ~ s e  GG w p p w k w o { vw k w ~ s s GiG w Gq0 sGGw ~ s  i vw k w w Gq0 sGiG w ~ s  vw k w m l k afi Gf sj {pp p f e f ~ s p p f e f ~ s ~ s p p p ~ s am fp l p p j p f e f ~ s p Gu0 Gq0 p p f e f f ~ s p Gu0 p Gq0 p f e f p ~s p o Guf e 0f ~ s { ~ s f Gu0 ~s
is a martingale under I , as we can check below: P

f ef f

ef

We then have the following pricing formulas: For a Simple European derivative security with payoff at time ,

f ge

(a) 114

Remark 9.4 Note that

Now for an American derivative security

Remark 9.3

More generally for

(b)

More generally for

is a supermartingale under I , P , , : (Lemma 2.28) (Lemma 2.29)

CHAPTER 9. Pricing in terms of Market Probabilities


() = 1.44 2 S2 (HH) = 16 1/3 1() = 1.20 S (H) = 8 1 1/3 S =4 0 2/3 0 = 1.00 2/3 () = 0.72 2 S2 (HT) = 4 S2 (TH) = 4 () = 0.72 2

115

1/3 S1 (T) = 2 () = 0.6 1 2/3

2() = 0.36 S2 (TT) = 1

(c)

is the smallest process having properties (a) and (b).

Example 9.3 (Radon-NikodymTheorem, continued) We illustrate the use of the valuation formulas for European and American derivative securities in terms of market probabilities. Recall that , . The state price values are shown in Fig. 9.2.

Compare with the risk-neutral pricing formulas:

Now consider an American put with strike price 5 and expiration time 2. Fig. 9.3 shows the values of . We compute the value of the put under various stopping times :
 4 q 'Eb' `'b '

(0) Stop immediately: value is 1. (1) If

 9 v b v qv v  T v  6 b v b 6 b T  bb v v s bb bb  q b  ` i s s b bb v b

, the value is

For a European Call with strike price 5, expiration time 2, we have

We interpret by observing that at time if occurs.

is the value at time zero of a contract which pays $1

W#

Figure 9.2: Showing the state price values

. The probabilities shown are for I , not P

Ws f

am ip l p p j
q

116
(5 - S (HH))+= 0 2 2 (HH) (5 - S2(HH))+= 0

+ (5 - S (H)) = 0 1 += 0 (H) (5 - S (H)) 1 1 1/3 (5-S 0)+=1 + 0 (5-S 0) =1 2/3 + (5 - S1(T)) = 3 (T) (5 - S (T))+ = 1.80 1 1

1/3

2/3 1/3

+ (5 - S2(HT)) = 1 + 2 (HT) (5 - S2(HT)) = 0.72 (5 - S (TH))+= 1 2 + (TH) (5 - S (TH)) = 0.72 2 2

2/3

+ (5 - S2(TT)) = 4 + 2 (TT) (5 - S2(TT)) = 1.44

(2) If we stop at time 2, the value is

We see that (1) is optimal stopping rule.

9.4 Stochastic Volatility Binomial Model


)#ri# H)ca2

Let I be the market probability measure, and assume P equivalent. Dene


#

v8 2 fi 2 #

~

cf #

aa

and for

c8 #

H82 #

Let

be the risk-neutral probability measure:

8Va'D

Let are

be the set of sequences of -measurable. Also let

tosses, and let

rs

Figure 9.3: Showing the values P I , not .


W#

for an American put. The probabilities shown are for

, where for each ,

. Then I and P

are

CHAPTER 9. Pricing in terms of Market Probabilities


i #Vcym2 c  i E4

117

We dene the money market price process as follows:


#

We then dene the state price process to be


i c2

As before the portfolio process is . The self-nancing value process (wealth process) , the non-random initial wealth, and consists of
i V) c28V))V)|c|fif)c )T

and the following processes are martingales under I : P


i E8

and

We thus have the following pricing formulas:


Simple European derivative security with payoff

at time :

American derivative security

The usual hedging portfolio formulas still work.

' & $ (%%# #  #  !  ") 

 

d a    

 #

# 

and

' & $ (%%# !  " 

Then the following processes are martingales under

V)

o)

Note that

is

-measurable.

)V

8V

#2T8

sm)'D

118

9.5 Another Applicaton of the Radon-Nikodym Theorem


Let be a probability space. Let be a sub- -algebra of , and let be a non-negative random variable with . We construct the conditional expectation (under ) of given . On , dene two probability measures

if for some , this is just the denition of , and the rest follows from the standard machine. If and , then , so . In other words, the measure is absolutely continuous with respect to the measure . The Radon-Nikodym theorem implies that there exists a -measurable random variable such that
# c'58 # 1 5 @v9 # 5 c'8%0

i.e.,

This shows that has the partial averaging property, and since is -measurable, it is the conditional expectation (under the probability measure ) of given . The existence of conditional expectations is a consequence of the Radon-Nikodym theorem.

1 5 @v9

A 0 G4

c F

E C %D

E C B8 # R 5

0 E 4%C

5 c6

1 P5

1 Qv5

Whenever

is a -measurable random variable, we have

A 1 5 B@v9

1 5 @v9

0 E C 5 42%D'8 #

5 8

0 5 76

0 4

0 4s~

H IF

Chapter 10

Capital Asset Pricing


10.1 An Optimization Problem

Consider an agent who has initial wealth as to maximize

Here, (BC) stands for Budget Constraint. Remark 10.2 If is any random variable satisfying (BC), i.e.,
c

then there is a portfolio which starts with initial wealth and produces at time . To see this, just regard as a simple European derivative security paying off at time . Then is its value at time 0, and starting from this value, there is a hedging portfolio which produces .

Remarks 10.1 and 10.2 show that the optimal obtained by solving the following Constrained Optimization Problem: Find a random variable which solves:

for the capital asset pricing problem can be

Equivalently, we wish to

119

c W US srpXVi

$ h

Maximize

Subject to

c W US g4fe

Maximize

c d

c 7

a 4b8

Y `E8

Remark 10.1 Regardless of the portfolio used by the agent,


W US XVT

and wants to invest in the stock and money markets so

is a martingale under I , so P

')

sr

x t
Equation (1.1) implies

w D V)

T V x t

) D

Ti

o8V sV

))

x t

x t
and

))

V)

#2m x t

f x t

)V t I' t x t x t '

t
)V

t
V)

t |

x yt

x yt

fi

cwGv ) ) ) t fi

sr

) s) W US t XVi

sr

s r c s #vdvr

)V

r c m8pdr )) r Tp)

c t 8iu t

120

Subject to

c Pq

$ h

There are

sequences

in . Call them

sr p

We can thus restate the problem as:

Maximize

Subject to

In order to solve this problem we use:

Theorem 1.30 (Lagrange Multiplier) If

Maxmize

Subject to

g'

))

Plugging this into (1.2) we get

sr

I'T x t

For our problem, (1.1) and (1.2) become


then there is a number

such that

solve the problem

. Adopt the notation

(1.2) (1.1)

c W US c W US x eXVT eXVe
c W US XVe

#a

c W US x eXVT oc %

W US QXVd

c c W U PpXVS
From (1.5) we have

x c
%

'oc

f

W US t o XV7' x t '

W U t t XVS x t
The function

t fa r t ' t

'

t qy

W US t t XV7' t ' x c x c

x c

'dc

V)

#'

x c

x t
Therefore,

Thus we have shown that if

CHAPTER 10. Capital Asset Pricing

solves the problem

Maximize Subject to

c W US pXVT

Theorem 1.31 If

Proof: Fix

We maximize

then

over

and so Taking expectations, we have and let Let be any random variable satisfying is maximized at and dene is given by (1.4), then : , i.e., solves the problem (1.3). (1.5) (1.4) (1.3) 121

#s#'D

V) ))

REfif

io|D

)) V)

REf

m & o|D k

m k h j if )( nl %'D
V)

h j if Rf

' D

)V

m4

gm

W US XVT

In summary, capital asset pricing works as follows: Consider an agent who has initial wealth and wants to invest in the stock and money market so as to maximize

and the optimal portfolio is given by

The optimal

122

is

h f ig

Since

so is a martingale under I , we have P , i.e.,

Chapter 11

General Random Variables


11.1 Law of a Random Variable
Thus far we have considered only random variables whose domain and range are discrete. We now consider a general random variable dened on the probability space . Recall that: is a -algebra of subsets of .
%c5 5 8 s vsH

P I is a probability measure on

11.2 Density of a Random Variable


 r )t~ p

123

v a H Q9

t ' t

%C

a 6

The density of

(if it exists) is a function

such that

a %8

where the probabiliy on the right is dened since in Williams book this is denoted by .

is often called the Law of

v sH pv2

8H

v Q

a 9

{ a |8

z a sI c8

Thus any random variable by

induces a measure

on the measurable space

H p

n'p r

i.e., 11.1)

is a random variable if and only if

is a function from

to

v u

a %22@|

A function is a random variable if and only if for every Borel subsets of I ), the set R
A Px8vR 8 a a R 2w )

q r p
, i.e.,

nu r p

t t

is dened for every

. (the -algebra of

(See Fig. dened

124

R B
.

{X B}

Figure 11.1: Illustrating a real-valued random variable

We then write where the integral is with respect to the Lebesgue measure on I . R is the Radon-Nikodym derivative of with respect to the Lebesgue measure. Thus has a density if and only if is absolutely continuous with respect to Lebesgue measure, which means that whenever has Lebesgue measure zero, then
H p

11.3 Expectation
~ #w r p

Theorem 3.32 (Expectation of a function of

) Let

be given. Then

which are true by denition. Now use the standard machine to get the equations for general .

t ' t

a @

a 8

aa

Proof: (Sketch). If

for some

, then these equations are

'sa

t ' t

t s t

y X' t XC

g 2w a

v t C

D' t

y X

H t I' t

CHAPTER 11. General Random Variables

125

(X,Y)

C y { (X,Y) C} x
F 

Figure 11.2: Two real-valued random variables

11.4 Two random variables


F v
be two random variables dened on the space . Then Let (see Fig. 11.2) called the joint law of , dened by measure on
s vsHp

induce a

that satises

f t t k v t yC C

F 's8a

d)

k f 4 C y C y t t

F 

We compute the expectation of a function of

k f y

is the Radon-Nikodym derivative of

with respect to the Lebesgue measure (area) on

in a manner analogous to the univariate case:

HQ v

 )tcr

f t ' t k

#p

k f

C yC

F 

'4i

F 

k f y

The joint density of

is a function

p Q v

F v

F Qv

cR 2i

n r

k f y

p v

F 

k f

v a H p~9

f t t k  t 4G%C C

'8

vd6f C

F fd'F ) a~ g r p F k f
F 2 B2i) F 4

k f XC R 8 t t
where

'd6

f t t k C C f aI k y 2aPI 2v F a F @

a 8

y
Suppose
F v

11.5 Marginal Density

126

has joint density

. Let

ga

k f

Therefore,

is the (marginal) density for

Suppose

11.6 Conditional Expectation

has joint density . Let be given. Recall that depends on through , i.e., there is a function ( depending on ) such that

R F )

How do we determine ?

We can characterize using partial averaging: Recall that . Then the following are equivalent characterizations of :

be given. Then

for some

(6.4) (6.3) (6.2) (6.1)

8H

v Q

a 9

k f 4' v8 H C C y t t

d8 q8 y

H XC

mH

a 9

v4 F

'i F

F qi

F 2 5 8i'@v9

E C

fbE F

C
v H Qa

a F 5 F 2 ovi'P5

n i Gi (  %vp  nX"|Xn
. Let

|

i) xq

f v EF V t C t t
DTcF V

f C t t k v t XG%C
8H p

F 82i

dF )a
'8

~f v C t t t

~ #@ r p ~f p r #Bd t

A function any function

11.7 Conditional Density

CHAPTER 11. General Random Variables

is called a conditional density for

given

 tr

and

(Here is the function satisfying

depends on , but

~f

Theorem 7.33 If

Proof: Just verify that dened by (7.1) satises (6.4): For

v Q

Example 11.1 (Jointly normal random variables) Given parameters: have the joint density
s

F fd'F )
and then replace the dummy variable

~f v XC t t t

c8f
The function

F 4fdF )

'd6

m Vj 'd f v XG%C C t t t

8 k f t

~f t
F 

has a joint density

k f

In conclusion, to determine

Notation 11.1 Let be the function satisfying

and (7.1) becomes

is often written as

does not.)

by the random variable , then

(a function of ), rst compute

: provided that for (7.2) (7.1) 127

'8

F )

9i g

 PC " %    `

X
. Therefore,

PC )

~ g
In the -variable,

q Qi

o e q

{ z  q

y 

i

o i)

(i 6

Qi 

{ z "

{ z V8|  B

P ~ PC  % Gi  PC
using the substitution

Gi

"% i i 

" Gi i
The exponent is

We can compute the Marginal density of

128

as follows

Conditional density. From the expressions

o )

we have

Thus

is normal with mean 0 and variance

is a normal density with mean

and variance

Gi  o o ) i)

9i e

CHAPTER 11. General Random Variables


From the above two formulas we have the formulas

129

Taking expectations in (7.3) and (7.4) yields

Based on , the best estimator of is . This estimator is unbiased (has expected error zero) and the expected square error is . No other estimator based on can have a smaller expected square error (Homework problem 2.1).

11.8 Multivariate Normal Distribution


Please see Oksendal Appendix A. , and the corresponding Let denote the column vector of random variables column vector of values . has a multivariate normal distribution if and only if the random variables have the joint density

Here,

where

is the variance of

2 )

V)

r 2 8W 2 r

y ~

y ~%

i.e. the th element of is if and only if is diagonal, i.e.,

. The random variables in

&

T8 T

8g c

y r P % c r 2

qs8

and

is an

nonsingular matrix.

is the covariance matrix

T

&

&

q Gi

q Gi %

)V

))

r s

|T8 r

`    

n `

"%

@ ~ ~! ( vXvf8 (" 'i4o


(7.3)

&

&

y 

V)

))

)r ) t t

y iR

(7.4)

(7.5)

(7.6)

are independent

5 24
i

y 2 r h r r
)V )

) 01 #!

5 4 3 2

) 01 #!

cR

(&$! '%"

&

 r y rr 2 r 8Tr t

Let denote a column vector with components in , and let have a and mean vector . Then the moment multivariate normal distribution with covariance matrix generating function is given by

If any random variables have this moment generating function, then they are jointly normal, and we can read out the means and covariances. The random variables are jointly normal and independent if and only if for any real column vector
P )) r 8GP)%

))

t '

V)

Vr

t R t

# & )) P )rs8% & 5

( r 5 Y & )V Y ! Y C Y C

&

! "

&

))

r )yi8

r y VPr

r 2 v2 y t )iT t

" D r r 2 y r iT' t

11.10 MGF of jointly normal random variables

 !

r 2 2 r ( " vX

'ir

fk t 8 t

and we have the formula from Example 11.1, adjusted to account for the possibly non-zero expectations:

r r "  92 |2

("



m P j   s  j  r r 2 r 2 vf2

r 2 f2

m s j m   s j 

r 2 2
r y r y 8e)iP|j%

Take
c%

Thus,

11.9 Bivariate normal distribution


130

in the above denitions, and let

a d F a d F a

V)

o m)a

V)

r a 2 F )a )~8'V

))

r F F T9Rfi2

6 B DC

)V

5 24

) 01 A!

5 4  2
i V)

) 01 A!

9&$! @'%"
f

Let be a collection of independent, standard normal random variables dened on where I is the market measure. As before we denote the column vector P by therefore have for any real colum vector ,

s PH

76

&

F 

)V

F )

&

If we know then we know

Dene the discrete-time Brownian motion (See Fig. 12.1):

12.1 Discrete-time Brownian Motion

Semi-Continuous Models

Chapter 12

, then we know

. Conversely, if we know . Dene the ltration

)V

r a Vx))Ha

d )V 8aTDv a F a r a r F a )~9)HF V) V) a r Va ~a F v r F )v)F

Theorem 1.34

a T`

Proof:

is a martingale (under I ). P

131

a a

, . We

F 

)V

2 r r
g

y F |fy2

( !

cf

s 2 r r

y a %ic2

( !

a 8V8

a dR )Td fi F Y

r H8G8 ! E

E F

F fi c8 f

7E

a 4 )Td F

a cR )ifi8

Theorem 1.35

Proof: Note that

which is a function of

Given parameters:

The stock price process is then given by

a a

y t t

a `

Note that

12.2 The Stock Price Process

Then

Use the Independence Lemma. Dene

132

, the volatility. , the mean rate of return.

, the initial stock price.

2 t

B k

is a Markov process.

alone. Figure 12.1: Discrete-time Brownian motion.

1 Y1

k 2 Y2 Y 3 3 4 Y 4

g fi fi

V)

U !G)f U Xfifd ! ! U W|f)

V)

#)g4'

U! VT

2 { | 2 r r   

y %iv fi

fif

r
133

P %! 82

P ! ! r P ! 2 r Q I! V F fy2

The other processes in the market are dened as follows.

F y2 z

XV7 W US

R mft

W US XV7y
Thus

m )fi

and

S ' R

fif

W US XVX

S @ R

Denition 12.1 Let


i V) 8

Money market process:

Discounted wealth process:

12.4 Risk-Neutral Measure

Portfolio process:

12.3 Remainder of the Market

CHAPTER 12. Semi-Continuous Models

Each

is

Wealth process:

Each

given, nonrandom.

is

-measurable.

-measurable.

be a probability measure on

76

is a martingale under

, we say that

is a risk-neutral measure. , equivalent to the market measure I . If P

t t

t t

134

Proof:

12.5 Risk-Neutral Pricing


Let

be the payoff at time , and say it is

-measurable. Note that

may be path-dependent.

Hedging a short position:

Sell the simple European derivative security


Receive

at time 0.
i V)

Remark 12.1 Hedging in this semi-continuous model is usually not possible because there are not enough trading dates. This difculty will disappear when we go to the fully continuous model.

12.6 Arbitrage
c

Denition 12.2 An arbitrage is a portfolio which starts with


#'#

(I here is the market measure). P

Theorem 6.37 (Fundamental Theorem of Asset Pricing: Easy part) If there is a risk-neutral measure, then there is no arbitrage.

#a

If there is a risk-neutral measure

, then

and ends with

satisfying

Construct a portfolio process

which starts with

and ends with

    f g Y ~ f  f ~ Y f

` a

f fi

g Y

a martingale under

, regardless of the portfolio process used to generate it.

Theorem 4.36 If

is a risk-neutral measure, then every discounted wealth process

is

t t t t

2 r

y o !

(v

2 r

y 

( !

2 r

f (v

U P!

F 2 8 Vfi%

   

! F 2

 fi

fif

2 r r

y ic fi

(v

f f

fif

)HV 2 r r

(  2 r |fy2 y F r ~! ( y F a Rfd62 !

s r 2 r

y %ic

a 2

( ! F v

s v%s V) r F )v)F

76

t t t

co

9H#a

'#c
'#c

'#

c'#a

c'#a

h fb Y iT # Y

, and let be the nal wealth corresponding Proof: Let be a risk-neutral measure, let to any portfolio process. Since is a martingale under ,

Suppose

` a 9'# ` c
'#

(6.1) and (6.2) imply

If

Therefore,

12.7 Stalking the Risk-Neutral Measure

This is not an arbitrage.

CHAPTER 12. Semi-Continuous Models

Recall that

, the market measure is risk neutral. If . . We have .

d Wy

cy

are independent, standard normal random variables on some probability space

. We have

, we must seek further. (6.2) (6.1) 135

g'

76 Y

5 j9

E C 5 %'8 Y

( ! F g %hD

g
h ) 01

( ! ! (

(v !

5 24

gr

F g ohD

( !



2 r d

( !

(v (v 2 r d 2 r r ! !

  

F ` 2 ! F
)V

fi

P F U f|fi7fF fiF

2 r

r r (

2 r

'

P U e'F |2

2 r

y F %i'fi2

under which We want a probability measure dom variables. Then we would have

fF

U P

! ! x!

(v (v (v

f f f

fi f

The quantity

where

136

is denoted and is called the market price of risk.

Then

Dene

Properties of :

Cameron-Martin-Girsanovs Idea: Dene the random variable

for all

j5

In other words,

is a probability measure. and are independent, standard normal ran-

5 8

t t



{ |

 g
r
#

r r z
p

f f


! ! ( ( !

! (

h h

 

gr

(v

! (v

 

g X
F i

gr  g

F g |#X

f f

F g ohD

h TF g



5 %%v9



g F

F 7

V)

g W

are independent, standard normal under

F 7 F
We show that

CHAPTER 12. Semi-Continuous Models

Verication:

Compute the moment generating function of

g W

F 7



almost surely.

is a risk-neutral measure. For this, it sufces to show that

: Independent, standard normal under I , and P

for every random variable

under

))

: 137

% cyH

E C

5 '8

Y
r

F g %hD f
V)

pi
!


(
F f7

t
F

g W

t t

F v

V)

r F )v)F

E s

{
s

q Xis r
r 2 r

2 r

E y2

( (

b! !

z
r

i U!

vxwqu 24 5 

2 ) 01

rt s U !

5 24

2 r

F F i F

h 2 h 2

) 01 A! ) 01
!

5 24

2 r

y y %isT%ia'

f f

P U e
y %ic

5 24

2 r

h 2 a 2

) 01 A!
!

P 2 r r

Stock price at time

Payoff at time

12.8 Pricing a European Call

138

is

is

since

. Price at time zero is

is normal with mean 0, variance , under

This is the Black-Scholes price. It does not depend on .


F f

q X

q X

Chapter 13

Brownian Motion
13.1 Symmetric Random Walk
Toss a fair coin innitely many times. Dene

Set

13.2 The Law of Large Numbers


We will use the method of moment generating functions to derive the Law of Large Numbers:

Theorem 2.38 (Law of Large Numbers:)

139

 r d"

almost surely, as

if if

H f

 c

'

r a

! r

! r
z ( !

c'

r 

! r r r S !
z W U XVS

! rr ! t r r ! !
{ X

! r

{ 4

5 24

Theorem 3.39 (Central Limit Theorem)

We use the method of moment generating functions to prove the Central Limit Theorem.

13.3 Central Limit Theorem

! r

! r
( ! h

which is the m.g.f. for the constant 0.

c '

W U XVS

Y r

z W U 4fS c'

W U XVS
which implies,

! r
z
f

) 01
! !

c c'

Proof:

Therefore,

Let

140

. Then

Standard normal, as (Def. of

 r d"

! T'

Y r

Proof:

(Independence of the

(LH pitals Rule) o ) s)

m
a

m j
a

'D

m j

13.4 Brownian Motion as a Limit of Random Walks

r !
(LH pitals Rule) o

! r

which is the m.g.f. for a standard normal random variable.

! r T'

Y r

r !

r S V!
r

r r r

S
r r

S S

'rd

! ! r ! r ! r S r ! { ! r ! r z t r r ! ! r t { ! r ! r XVS z WU

W U XVS

! r

z W U 4fS c'

W U XVS

so that,

Therefore,

Y r

Let

CHAPTER 13. Brownian Motion

Let

be a positive integer. If

If

is not of the form , then dene . Then

j a

Here are some properties of


is of the form , then set

by linear interpolation (See Fig. 13.1). (LH pitals Rule) o 141

f s s )

cs#p m c#

m m

s s s

a 8 j S ' a j

j a

p Bo m
#) s) m )D

j a a j 8

S' R

h h

f s f s

) )

S ' R

m m

a j

)D

j a

p BVD m

j a

m j
a

142

Figure 13.1: Linear Interpolation to dene

Also note that:

Properties of

Properties of

is a continuous function of .

` dmy m j

a j T'o m

j a

and

To get Brownian motion, let

k/n

in

(Please refer to Oksendal, Chapter 2.)

13.5 Brownian Motion


 r
are independent. . (Approximately normal) (Approximately normal)
VD D m j a s j a t

(k+1)/n

8)

T' a a

))

a r a r mRs T'8| 9F V)

r ~

A random variable properties:

13.6 Covariance of Brownian Motion

S e A' F '

F v

c V)

r F )v)F

F R t t t

T'RPF a a

WPW A

y a

CHAPTER 13. Brownian Motion

(, F,P)

1.

g'#a

2.

is a continuous function of ;

D a

3.

Let

then

and

has independent, normally distributed increments: If

be given. Then and are jointly normal. Moreover,


) a

Figure 13.2: Continuous-time Brownian Motion.

(see Fig. 13.2) is called a Brownian Motion if it satises the following

are independent,

are independent, so

B(t) = B(t,)

and

D a

) a

7f

f j i f )h ) a sV7'y8)V7 g f a a f a r B)7ad P)7T'y88V7 gyvV7 f f a a f a a f a A'sy8 a a gy S@ ofT's)ha8 f f a S @ R R g)h a ) 7Hy a

f h

7f

a a a ) dsV 'y 8 c

7f

7f f g g

143

7q D)

V) r 88V88)88V888)8 V)

r r ps o  s V) sps n o a a n V) 8v a Io8sv n a r a r 8)8)8V8)V8)8)8V)8V8)8)8V)8V8)8)8V)8V8)8)V8)8V8) v r a a V) r v r a a a o  vRP V) 8vir s is r a nn a a r a a Ioo n


P

) a W

))

a r a 8 8 WPW2

V)

r ~

Thus for any

144

(not necessarily

mklfTD)7fa a Wf

Required properties:

13.8 Filtration generated by a Brownian Motion

is jointly normal with covariance matrix

` f

be given. Then

13.7 Finite-Dimensional Distributions of Brownian Motion

Let

For each ,

H vd H t f QDo7Xp mvQV7% f a f a H HQ v pHtf y D T' a )) m aT' r a )DT' a a a W#Q%rT~WPW rr r ya

in . Do this for all possible numbers required by the -algebra properties.


y

is

This

Here is one way to construct

are independent of

For each and for

contains exactly the information learned by observing the Brownian motion upto time . is called the ltration generated by the Brownian motion. . -measurable, . First x . Let , the Brownian motion increments ), we have and and . Then put in every other set be given. Put the set

D ) D

7q

t t

2r 2

r s s c


     

) V

f )7 ()fs g r 'P)7'D8 r f a a ' r h r ~! ( )7f S g  f r f a R a g ( f  VX r g 's)hy8hd ! s)7   f X| r g r 's)hy8ahd ! ( V7f c f a g

7f

) j )

f h

f f G)Xs r g r

a a a 's) d) TD 8 d

f h

r g r

a y %

h g

( !

Theorem 9.41 Let

2g

a V D

a V d B) D g a a

a a a a y) )V dfs) y 8 gp) Vy

7f

7f

7f

7f

f h

7f

7f

Theorem 9.40 Brownian motion is a martingale.

13.9 Martingale Property

CHAPTER 13. Brownian Motion

Proof: Let

f WG

Proof: Let

13.10 The Limit of a Binomial Model

is a martingale.

Consider the th Binomial model with the following parameters:

7f

g h

Up factor. (

f cBV7 )y

f WG

t s t

Down factor. be given. Then be given. Then be given. Then ). 145

t t t t t

m j

r )

v r ) 2 Pd

}~ r x %s w

r 2

2 r 2 r

w  r
r

2 r

z
2

}~ %s 2 | w rx r

W U XVS

2 r 2

z {c

W U D4fS

2
W US D4f

2
'

W U DXVS

W U DXVS

2 r

W US 'DXV"

y

m j

W U XVS

x w yt W
r r 2

r t

'

W U t DXVS

ds

m j

m j

2p Hp

r r

'dP ' p

u u

' g'dP s

uu u F

u u

Let rst

146

denote the number of tosses. Then

in the rst

tosses, and let

In the th model, take

which implies,

steps per unit time. Set

. Let

denote the number of

for some , and let

v m Ds b j

'

2


v m Ds b j

m j

m j

Theorem 10.42 As

where is a Brownian motion. Note that the correction martingale.

2 r

a 2 D `

dP

Under

, the price process

, the distribution of

 r d9

As so

Proof: Recall that from the Taylor series we have

, the distribution of

is a martingale.

approaches the distribution of

2 r

a D 2

W U 4fS

 r

converges to the distribution of

is necessary in order to have a

. in the

CHAPTER 13. Brownian Motion

147

B(t) = B(t,) x t

(, F, Px)
Figure 13.3: Continuous-time Brownian Motion, starting at

13.11 Starting at Points Other Than 0


(The remaining sections in this chapter were taught Dec 7.)

For a Brownian motion that starts at , denote the corresponding probability measure by (See Fig. 13.3), and for such a Brownian motion we have:

Note that: If , then

puts all its probability on a completely different set from I . P

13.12 Markov Property for Brownian Motion


We prove that Theorem 12.43 Brownian motion has the Markov property. Proof: Let

be given (See Fig. 13.4).

Independent of

-measurable

I%7

~aW

Q7a

l l Q7aX~ W7a@  %h

The distribution of

under

is the same as the distribution of

under I . P

'  a7

@e T ah

H~sl7a7

For a Brownian motion

that starts at 0, we have:

m7 HQ7ah hQ7

Hm7 h@' a ' G%haXsha' IG 7


Use the Independence Lemma. Dene
same distribution as

In fact Brownian motion has the strong Markov property.

Then

148

Example 13.1 (Strong Markov Property) See Fig. 13.5. Fix

' e ' 'l % l   @ ~pX

sl7m7

Then we have:

B(s)

Figure 13.4: Markov Property of Brownian Motion.

restart

and dene

s+t

'e yw ya

3
is a martingale, and

y ymC 3 a  "T

"

  D 7 yh H~s7m7   D %7m 7 %7m7 Q7 H7m7   D 7 Ha7 7 C p~CIl ' D 


to in time , and

Let be the probability that the Brownian motion changes value from let be dened as in the previous section.

D 

Fix

Fix

13.14 First Passage Time

CHAPTER 13. Brownian Motion

13.13 Transition Density

. Then . Dene

Figure 13.5: Strong Markov Property of Brownian Motion.

restart +t
149

 D

' F m Fm Fm F m lF m m F

H F m
 C 

 3

6 7

y s

m ' G y
   ( )

& '

'e
0

$ # %

We have

150

if if

  

yw ywa y y m y
   

l


We use the Reection Principle below (see Fig. 13.6).


A '

6 7

Conclusion. Brownian motion reaches level is innite.


4 @


8 9 3 4 5

( )

& '


1 2  

y
! "

Let

Let

to get

$ #

Letting

Differentiation of (14.3) w.r.t.

Let

in (14.1), using the Bounded Convergence Theorem, to get

, we obtain

. We have the m.g.f.:

, so

yields

with probability 1. The expected time to reach level (14.4) (14.3) (14.2) (14.1)

1 B3

6 7

@H
4 @  U

4 @


Laplace transform formula:


d V V d eV b ` E T E c a

X Y

V W

p ' ' H ' H


`  U E T R S 


E G

6 E

CHAPTER 13. Brownian Motion

Using the substitution

6 F H PI

then Density:

which follows from the fact that if

Figure 13.6: Reection Principle in Brownian Motion.

we get

t shadow path
151

Brownian motion

152

Chapter 14

The It Integral o
The following chapters deal with Stochastic Differential Equations in Finance. References: 1. B. Oksendal, Stochastic Differential Equations, Springer-Verlag,1995 2. J. Hull, Options, Futures and other Derivative Securities, Prentice Hall, 1993.

14.1 Brownian Motion


x

2.

is a continuous function of ,
y Y  G   f

3. If

, then the increments


f

are independent,normal, and

14.2 First Variation

153

Quadratic variation is a measure of volatility. First we will consider rst variation, function .

)`

e au a m yT h au a m

y Y

p P

p 7

1.

Technically,

g wvu t

r s

p q

h i

(See Fig. 13.3.) Brownian motion,


g c

aX a %Q aX a Q%T e j y a yj W  m
,

is given, always in the background, even when not explicitly mentioned. , has the following properties:

~a a
U

, of a

154

f(t)

t2 t 1
U

Thus, rst variation measures the total amount of up and down motion of the path. The general denition of rst variation is as follows:
i y f

Denition 14.1 (First Variation) Let

be a partition of

The mesh of the partition is dened to be

Suppose is differentiable. Then the Mean Value Theorem implies that in each subinterval there is a point such that

X %

ee e gf e

d 7f

)`

We then dene

For the function pictured in Fig. 14.1, the rst variation over the interval
U U U U U U 5 7f B)`

@ ' ~  @~ '   y
U

@U

Figure 14.1: Example function

. is given by:

Q%T y%Q%

 G

c7f 

, i.e.,

i 

'e

X
  i U

e e
 i f h y U   i U


m oU l


 f U U




Denition 14.2 (Quadratic Variation) The quadratic variation of a function on an interval is


  i U

e
 P i U

In particular, the paths of Brownian motion are not differentiable.


r

m
p q m l

p q7 p


m l

yT
f

@
U f j f y h

ee e gf e

eegf e e

eegf e e


m eU l

 j h f h y


y U U

f y h

eenf e e


m eU l

@
U f f y h

kf

ee e gf e


U 5 7f )` h y

f j h

Then


y U U

or more precisely, Theorem 3.44 and

Remark 14.1 (Quadratic Variation of Differentiable Functions) If is differentiable, then , because

14.3 Quadratic Variation


and

CHAPTER 14. The It Integral o

155


f f

u u

yj

{  z

j h

% e
y


| 

s i P

| ts

X X e e T C e % X A
f h } ~ y f h y P  s f h

{  z

m
y y s

yj X
P  P

j h


y P 


This has expectation 0, so
s

C X A C as a X A

T
P P


f h y

ee e gf e

e F
s f


. To simplify notation, set
ts

f u

aX a

y


be a partition of Proof: (Outline) Let . Dene the sample quadratic variation

yQ%Q f
y f

Then

A
f

For

Consider an individual summand

We want to show that

156

, the terms

and

e F
y y

are independent, so

(if

is normal with mean 0 and variance

, then

Thus we have )
u

{  z

w x

CHAPTER 14. The It Integral o




157

Remark 14.2 (Differential Representation) We know that

We showed above that


{  z

which we can write informally as

14.4 Quadratic Variation as Absolute Volatility


i

On any time interval

, we can sample the Brownian motion at times


and compute the squared sample absolute volatility


h y

This is approximately equal to


l

As we increase the number of sample points, this approximation becomes exact. In other words, Brownian motion has absolute volatility 1. Furthermore, consider the equation

This says that quadratic variation for Brownian motion accumulates at rate 1 at all times along almost every path.

'e

H aX a7

Q%T

' ~a ~a

 G

'

When

is small,

e H mu m7 % X C e F ~' aX a7

is very small, and we have the approximate equation

yj e F H mX a7 a

T
P 

ee e gf e P P

{  z

As

, so

158

14.5 Construction of the It Integral o


The integrator is Brownian motion following properties: , with associated ltration

We want to dene the It Integral: o

This wont work when the integrator is Brownian motion, because the paths of Brownian motion are not differentiable.

14.6 It integral of an elementary integrand o


i y f

Let

be a partition of
f

Think of

as the price per unit share of an asset at time .

m m

The functions

and

can be interpreted as follows:

Assume that is constant on each subinterval elementary process.

y  y

, i.e.,

(see Fig. 14.2). We call such a

Remark 14.3 (Integral w.r.t. a differentiable function) If we can dene


U c f c U c

Q%T y

y Y

 m y

 S 

 G

@~

y%Q%

~y

2.

is square-integrable:

1.

is

The integrand is

, where (i.e., is adapted)

-measurable

is a differentiable function, then

3. For are independent of


 T  

, the increments .

mu m y%Q% mu a au m

3 ~ %%j ~a

2.

is

-measurable,

Q7

1.

every set in

is also in

~a

, and the

an

a 
c c f

e c c

a 
c c

 m y
c

~
Linearity If Adaptedness For each is

 

 


  t

mu~a % 3 y aX a y ~ au m 3 mX a y 3 as a mXa y 3 as a D m ~a ~ % 3 %Q%


  y y y f y h    f f f f f

CHAPTER 14. The It Integral o

0=t0

( t ) = ( t ) 0

Figure 14.2: An elementary function .

Think of

Think of trading date

as the number of shares of the asset acquired at trading date .

Then the It integral o

14.7 Properties of the It integral of an elementary process o

f f

y y

In general, if

then ,

y
f

can be interpreted as the gain from trading at time ; this gain is given by:

as the trading dates for the asset.

t1

( t ) = ( t 1 )

-measurable.

( t ) = ( t 2 )

t2

( t )= ( t 3 )

t3

t4 = T

and held until 159

mX%7m H as h%7 m m  H aX m7% y


y y y

3 3 Q7 %7
h

H
y

mX m7% y as a7Q y
h

aX~a y aX m y D mX m y aX m y ~
h y y y y f y h c


 

160

Figure 14.3: Showing and in different partitions.

and

 mH y
c c f

Theorem 7.45 (Martingale Property)

Proof: Let be given. We treat the more difcult case that and and subintervals, i.e., there are partition points and such that Fig. 14.3).

aX~a C y mX m y
y y f y h

We prove the martingale property for the elementary process case.

We compute conditional expectations:

 r

Write is a martingale. Martingale is a martingale.


t l s t l+1

.....

k+1

are in different (See

 e % C ~ as a7
y y y y f y h

m
y s  y f & y h s y s   s  y f

y h

~ "
y

y y
y

jy

as a y  "
f y h  f

I %h H auQ D m % y %7 m %h H mu a a 3 Q7 H as a7Q y Q7

Since the cross terms have expectation zero,


f y s y h h


y y

s 

H aX~a7QC y
 P

~ mu m7% y
y

Q7

Each

Proof: To simplify notation, assume

Theorem 7.46 (It Isometry) o

These rst two terms add up to

CHAPTER 14. The It Integral o

has expectation 0, and different

. We show that the third and fourth terms are zero.

, so

are independent. 161

162

path of

path of

0=t0

t1

t2

t3

t4 = T
i

14.8 It integral of a general integrand o


Fix

. Let be a process (not necessarily an elementary process) such that


 p

is

-measurable,

Proof: Fig. 14.4 shows the main idea.

In the last section we have dened

for every . We now dene

~a ~

Theorem 8.47 There is a sequence of elementary processes


such that

Figure 14.4: Approximating a general process by an elementary process

, over

yT ' 3 y

m~

~a 3

y
f

 S 

@~ ~ 3
f

CHAPTER 14. The It Integral o

163

The only difculty with this approach is that we need to make sure the above limit exists. Suppose and are large positive integers. Then

14.9 Properties of the (general) It integral o

Here is any adapted, square-integrable process.

Linearity. If

and

Martingale. Continuity.

is a martingale.

is a continuous function of the upper limit of integration . .

Example 14.1 () Consider the It integral o

We approximate the integrand as shown in Fig. 14.5

It Isometry. o

a 

 mH y

then

a 

e c c

 m y

Adaptedness. For each ,

is

 a  y

-measurable.

which is small. This guarantees that the sequence

@ y ~ H @ ~y ' y y @ ~ ~a ~ H
has a limit.
f c f y y y f f

(It Isometry:) o

"

~ " ~ ~

{ z

h @


h 

7 h

@ h 

@ j

h @

h 


7 h @ 7 h @

h @

pp

nr

gr

7 h @

pp


u7u

y


164

Figure 14.5: Approximating the integrand

a
c

T/4

2T/4

if if

uvD

if

D
u

so

We compute To simplify notation, we denote By denition,

3T/4

with

, over

 a  y
c f

 mH a
c c

 
c U U

T
U


) s h

Theorem 10.48 (Quadratic variation of It integral) Let o

T  m m
c c f


c U U f c


c U U U


7 h @


7 h

Remark 14.4 (Reason for the

Therefore,

@

The extra term there, because

In contrast, for Brownian motion, we have

gr

7 h 

Let

or equivalently

CHAPTER 14. The It Integral o

but

Then

~
m l

14.10 Quadratic variation of an It integral o


and use the denition of quadratic variation to get

comes from the nonzero quadratic variation of Brownian motion. It has to be term) If is differentiable with , then 165

(It integral is a martingale) o

j h

j h

yyy    e %
y m l P P  T

yy X %Q 7aX ha % h  7 I
y y   h m l m l y y P

7 7
y  f 

Q
  m l m l m l

h y

~y

7mu 7m % y  mQ y  %%j y%Q% y %Q% f ' ~a ~a


f y f

'~

y
c

This holds even if is not an elementary process. The quadratic variation formula says that at each time , the instantaneous absolute volatility of is . This is the absolute volatility of the Brownian motion scaled by the size of the position (i.e. ) in the Brownian motion. Informally, we can write the quadratic variation formula in differential form as follows:

Proof: (For an elementary process ). Let for . To simplify notation, assume

~ ~

Compare this with

166

be the partition for , i.e., . We have

It follows that
 P

Q y


so Then Let us compute . Let

be a partition

Chapter 15

It s Formula o
15.1 It s formula for one Brownian motion o
We want a rule to differentiate expressions of the form , where is a differentiable were also differentiable, then the ordinary chain rule would give function. If

which could be written in differential notation as

However, is not differentiable, and in particular has nonzero quadratic variation, so the correct formula has an extra term, namely,

This is It s formula in differential form. Integrating this, we obtain It s formula in integral form: o o

Remark 15.1 (Differential vs. Integral Forms) The mathematically meaningful form of It s foro mula is It s formula in integral form: o

167

H m7

H m7

h

U c c H U U

p 'D H~ah fmH~ah

m7
H

HHa7

mHm7 ' Hm7

f a H a7 D H m 7 H~a7

f a H a7 H m7 H~a7

U c c U U c c U U

@ m7
U U

H~ah

Ha7

~a

~a

168 This is because we have solid denitions for both integrals appearing on the right-hand side. The rst,

is an It integral, dened in the previous chapter. The second, o

is a Riemann integral, the type used in freshman calculus. For paper and pencil computations, the more convenient form of It s rule is It s formula in differo o ential form:
U U U

There is an intuitive meaning but no solid denition for the terms and appearing in this formula. This formula becomes mathematically respectable only after we integrate it.

15.2 Derivation of It s formula o

In this case, Taylors formula to second order is exact because


X

is a quadratic function.

C ah

mX

C aX

y Y

Fix

and let

be a partition of

. Using Taylors formula, we write:

In the general case, the above equation is only approximate, and the error is of the order of . The total error will have limit zero in the last step of the following argument.

% 7

Q s h f% 7 X h 7 7

U @

a @C as m m m H m7 mX m C ah H m7  H m7 H m7 y %%Q

Let

be numbers. Taylors formula implies

' 7

Consider

, so that

~a m7 @H~a7 ~a H~a7

 mH a7
c T @U f

H a7

Hm7

7

 a 
c


T

~a

'
U

~a @H~ A @  G ' Hm
U U U U U U


U U

~ay

7
G c f

 m 7 f a H m7 fy m m
U c c m l

We let

15.3 Geometric Brownian motion

This is It s formula in integral form for the special case o

CHAPTER 15. It s Formula o

Denition 15.1 (Geometric Brownian Motion) Geometric Brownian motion is

f f

H m7 H m7
U U

where

Dene

and Geometric Brownian motion in integral form is

@H~

Thus, Geometric Brownian motion in differential form is


H~a
U

According to It s formula, o Then so and to obtain

are constant.

169

170

15.4 Quadratic variation of geometric Brownian motion


In the integral form of Geometric Brownian motion,

is not differentiable. It has quadratic variation

15.5 Volatility of Geometric Brownian motion




on

is

As , the above approximation becomes exact. In other words, the instantaneous relative volatility of is . This is usually called simply the volatility of .

15.6 First derivation of the Black-Scholes formula


Wealth of an investor. An investor begins with nonrandom initial wealth and at each time , holds shares of stock. Stock is modelled by a geometric Brownian motion:

@H

y Y

9

Fix volatility of

. Let

be a partition of

. The squared absolute sample

Hm

Thus the quadratic variation of we write


is given by the quadratic variation of


 a 

@H

y Q%Q

is differentiable with

. This term has zero quadratic variation. The It integral o

~
f

the Riemann integral


`

a 
c


G c

. In differential notation,

9v  1

CHAPTER 15. It s Formula o

171

can be random, but must be adapted. The investor nances his investing by borrowing or lending at interest rate .

The differential of this value is


G

satisfying the terminal condition


f 

If an investor starts with and uses the hedge for all , and in particular, .


F H~ ~ H y 7 

  C

'

In conclusion, we should let

be the solution to the Black-Scholes partial differential equation

'

H F

(where

and

) which simplies to

'

But we have set we obtain

, and we are seeking to cause

to agree with . Making these substitutions,

 2

'

Equating the

coefcients, we obtain:

H~

To ensure that coefcients, we obtain the




for all , we equate coefcients in their differentials. Equating the -hedging rule:

, then he will have

H F @ w F

A hedging portfolio starts with some initial wealth time tracks . We saw above that
 2 

and invests so that the wealth

' @ @ f3 f

Value of an option. Consider an European option which pays the value of this option at time if the stock price is option at each time is

at time . Let denote . In other words, the value of the

m ' D w a@HF @V w F a H @H~ ~ '%~ ~ ~F ~ H~ F


Risk premium


~ H

H~

@ @ H

 2

F


Let

denote the wealth of the investor at time . Then

at each

y
a

y
a

jb

jb

y ~y y y y y y @ ~y 3 ~y
jb

jb

jb

@
!

y
d T

l H~y l y '
d c d c c f d

H y
c f G

y ~y
c f d

 d

n

n

H y 3 y a  3 3 a ' '~y '~y ~a ' @y 'H~y ~a 3 'H~y ~y ~a y ~y ~y y f ~y y H3


@U

U U

h a  3
c U

H y
G c c u

a y

@HH3 l ~y
G

We apply It s formula to compute o

where

15.7 Mean and variance of the Cox-Ingersoll-Ross process

172

The Cox-Ingersoll-Ross model for interest rates is

and

are positive constants. In integral form, this equation is

. This is

Hy
c

~
T

3 y

If

We solve for

Integration yields

which implies that

Differentiation yields

Taking expectations and remembering that the expectation of an It integral is zero, we obtain o

The mean of

, then

. The integral form of the CIR equation is

for every . If

, then

, where

exhibits mean reversion: . We obtain

~ mu m Q%Q mu a %QT
is

b b

~y @
d

y 
d

yy
G c


d


T

 y

a 
c

c @


G c

Variance of

CHAPTER 15. It s Formula o

. The integral form of the equation derived earlier for

~
T

y
G c

Using the formula already derived for algebra we obtain

Denition 15.2 ( -dimensional Brownian Motion) A -dimensional Brownian Motion is a process


jb

y
b

Associated with a -dimensional Brownian motion, we have a ltration

H~ y%Q% ~ h ~a
H


jb

jb d

y y A  y

jb

 T ~ ~y
d

with the following properties:

15.8 Multidimensional Brownian Motion

d 7

{ z

d 7

@
which implies that

Differentiation yields

Taking expectations, we obtain

Each

is a one-dimensional Brownian motion;

y
C

If

, then the processes

and

For each , the random vector

For each

 T

are independent of
 

, the vector increments

and integrating the last equation, after considerable

-measurable; are independent.

is

such that 173

h y

{  z

T e C X e Q X s Q X u
y  y  

{ z

% X % X u X u s
y  y 

yT
f

f h

The increments appearing on the right-hand side of the above equation are all independent of one another and all have mean zero. Therefore,

X Q X
y

i f

 %Q%
to be
y v

y e

y T ~
. For
v i y

' ~
is a one-dimensional Brownian motion, we have the informal equation

15.9 Cross-variations of Brownian motions

174

Because each component

Proof: Let of and on

However, we have:

Theorem 9.49 If

We compute


h y j h y {  z

But expectation

All the increments appearing in the sum of cross terms are independent of one another and have mean zero. Therefore,

and . It follows that

&

As

, we have

. First note that

be a partition of

, so

converges to the constant

{  z

are independent of one another, and each has

, dene the sample cross variation

Uf

F iF U

  'e f H H f H " F 
U U

F H H F
c  c U U  

F F U

 

 

a@

D H
i 

~  @ @ % H 'w @ @ @ H H 'w
and and
   c 3 c    c c H  3 c 


 

' @e
 3  3

15.10 Multi-dimensional It formula o

To keep the notation as simple as possible, we write the It formula for two processes driven by a o two-dimensional Brownian motion. The formula generalizes to any number of processes driven by a Brownian motion of any number (not necessarily the same number) of dimensions.

Such processes, consisting of a nonrandom initial condition, plus a Riemann integral, plus one or more It integrals, are called semimartingales. The integrands o and can be any adapted processes. The adaptedness of the integrands guarantees that and are also adapted. In differential notation, we write

l
c c c  f f

 c ~   H  F F 
c

Let

In integral form, with is

Let be a function of three variables, and let have the corresponding It formula: o

Given these two semimartingales

CHAPTER 15. It s Formula o

and

be processes of the form

and

 

 

where
U U

, for

as decribed earlier and with all the variables lled in, this equation

, the quadratic and cross variations are:

, and

be semimartingales. Then we

175

176

Chapter 16

Markov processes and the Kolmogorov equations


16.1 Stochastic Differential Equations
Consider the stochastic differential equation:

Here and are given functions, usually assumed to be continuous in chitz continuous in ,i.e., there is a constant such that

for all
f

The solution process will be adapted to the ltration generated by the Brownian motion. If you know the path of the Brownian motion up to time , then you can evaluate .

Example 16.1 (Drifted Brownian motion) Let


2

be a constant and

, so

177

If

is given and we start with the initial condition

Wl

V y ~F  ~F Q7a %7F 7 HQ7F h F

Let be given. A solution to (SDE) with the initial condition satisfying


f 

is a process

y F

mF 'H F

(SDE) and Lips-

D 


D  
c f d  

 d 

F


178
then

Given the initial condition

16.2 Markov Property

We have the Markov property


 7f

In other words, if you observe the path of the driving Brownian motion from time 0 to time , and based on this information, you want to estimate , the only relevant information is the value of . You imagine starting the at time at value , and compute the expected value of .
f

F H  H F p H F

the expectation of condition

, given that

. Now let

F  H F

H F

H F F

Let

be given and let

be a function. Denote by

be given, and start with initial

Again, to compute the differential w.r.t. , treat


c 2

and

as constants:

V y

the solution is

Example 16.2 (Geometric Brownian motion) Let and

3V

3V y

To compute the differential w.r.t. , treat

and

as constants:

be constants. Consider

V


2'u n

f I Q h v s eQ f IG

Q h I v s IG r kRTeQ f

i Ff s

v jG Y ~}wz

i o c

q S VW

z i@D

cwQ h v s eG F I I h Q h xv s PG I I v f z

v S { z D x Q h IG b v Q s IG |k@yiRedHwedb a tu Q h P7HsQ s P7b IG b v IG q S Q s IG VTnPgE q S uQ h PHE IG

f r Q h I v s IG tRTPQ f

i f s

v jG Y Q Q h IG b v Q s IG b VqekRe7p0tedG

o i c

Q IG b D Q IG Re7@'RegE

Y I D Q IG E j S Q IG E mXlRPgkuTRPgD

Q q IG RtP(U

h I v s RTI

s I

h Q h gv s eG I I f Q Q I eeCQ h v s edWG IG U Y qG v v

@Q h v s PG I I Q s IG tegE

S Q q s I h I TFR0TtiReG y q S Q h IG rpiRegE

Q h I v s IG U Y tRweu7q

f Q IG b D Y I D U S Q IG E Redca`XWVTRPHFD

A9 ) & 0('(

9 9 9 @

'g 4

$% #

 B C

% " $ # 

) & 0('

 

g
Y

Denote by

g

The Markov property says that for

2 31

!   "  

the density (in the

16.3 Transition density

CHAPTER 16. Markov processes and the Kolmogorov equations

variable) of

Note that depends on and and dont depend on .

h RI

!6 875 



or equivalently,

i f s

v jG ~dv

 Fi S z

The derivative is The random variable with initial condition


Q q I RtPG y i

@D

S z '@D

and we are making the change of variable

Example 16.4 (Geometric Brownian motion) Recall that the solution to the SDE

Conditioned on , i.e.,

Example 16.3 (Drifted Brownian motion) Consider the SDE

, the random variable

, is Geometric Brownian motion:

only through their difference

has density

, conditioned on

or equivalently,

i " i

z @D @D

and for every ,

is normal with mean

. This is always the case when

Q h xv s eG I I

. In other words,

and variance

179

) i(

 (

) i

Yg

g

0
Y

 (

iYg

& (T

& '

 '
j

&

'g

i@f B QRIgvpG f s

v Q I v G j wRxa8@dY i f s

Q IG RegE

e 6

I v a Q IG RegE

Q I v B Rgd8G f

Y Q I v G j qRp@pY

Q v Q G E t exgw|HG e(w S B RPG Q I f q 0@D i f s # S q 7@D i f s Fe q q e  i qi q (C(wS Q v CxpG y h d# gwv

Q I v Ra8G f Q I v i RxdG f i Ff s

i cf s

v Q I v G j 0Rp@Y Y Q I v G j qRa8@aY

D Q q I @FR0WetPG

f "@D

h f

I Q h gv s eQ f IG

i @f s

v jG v

Fe f

16.4 The Kolmogorov Backward Equation

i e 6 (w( wv i Q IG E eRegdS 9 9 9 9 CxH|8gG (w 6 Q v Q G E h t

Ixvd

g#

S Q G cg

I v p I v xd

S Q v Q G E gCxH|HG t

Q h I v s I @ RTPG Q h I v s I kRxTPG v

i S S D Q q s I h I W@wtRPG

v { D x Q s IG cx'egE a tu

Therefore,

where

Using the transition density and a fair amount of calculus, one can compute the expected payoff from a European call:

Therefore,

180

Consider

The variables

and let

and

be the transition density. Then the Kolmogorov Backward Equation is:

in

are called the backward variables.

'g

In the case that and their difference becomes




are functions of alone, . We then write

depends on rather than

and

only through , and

(KBE)
 (

 g

(KBE)

 !   

 '
j

) i

& l

& '

 

&

k y

f q f Fee

i f s f i i Y

q j iuS y

h f

@CQ f

i@f s

v jG v

i @ y

S Q q TFR0TqG y

f Q IG b D Q IG Re7@'RegE

Y I D Q IG E j S Q IG E mXlRPgkuTRPgD

f f B f Q U v q v e"@HwG Y y v Q U v q v G "@wHU y 6 S

S t y s f Y U y

f S f Y v f Q U v q v ecpgG y y f q q t S Y S f U v q v @Hw U v q v @pHp y y y y q f S S U v q v @wH y y y f f B Y Y v 6 S Q U v q v G "@pHpU Q"@UgqpvG v y @ h v v f CQ"@UHvqpvG feQ"@UgvqpvG @ F h S v S f eQcUgvqpvG (y y @ F S Q q TFR0`G y f h f Q Q U Y qG v Ce"@aWG v Q IG b D Y I D U S Q IG E Re7@d`XWVTRPHFD

Let

16.5 Connection between stochastic calculus and KBE

It is true but very tedious to verify that satises the KBE

Example 16.6 (Geometric Brownian motion)

This is the Kolmogorov backward equation.

Therefore,

Example 16.5 (Drifted Brownian motion)

CHAPTER 16. Markov processes and the Kolmogorov equations

Consider

be a function, and dene

(5.1) 181

& i

&

& & i

&

&

w w

&   & 

 w

&
j

   A  R

9A 9 9

)i  !   

A9 9 9 9 B

B 9A 9 9 9

!6  6

 

!6

9A 9

 '

!     

 B

!6

9A

2 31

 B

9 9 9 @



 l

 (

 &

( i

 '

) & 

! !

# # # # !

 '
j

i 
j

& 

 '
j

& ('(

 '  '
j

& l

Theorem 5.50 Starting at

0
j

It s formula implies o
   

# 0

where

2 x1

Let

5

so

Proof: According to the Markov property,

Therefore, the Kolmogorov backward equation implies

182

be an initial condition for the SDE (5.1). We simplify notation by writing

. Then

, the process

satises the martingale property:

(Markov property) rather than



CHAPTER 16. Markov processes and the Kolmogorov equations


In integral form, we have
1

183

Thus by two different arguments, one based on the Kolmogorov backward equation, and the other based on It s formula, we have come to the same conclusion. o Theorem 5.51 (Feynman-Kac) Dene

The Black-Scholes equation is a special case of this theorem, as we show in the next section. Remark 16.1 (Derivation of KBE) We plunked down the Kolmogorov backward equation without any justication. In fact, one can use It s formula to prove the Feynman-Kac Theorem, and use o the Feynman-Kac Theorem to derive the Kolmogorov backward equation.

16.6 Black-Scholes
) i

the solution is

x

 

With initial condition

& 

& 

 @'

Consider the SDE


(&

) i

 T0

and

 0

 7mip'

Then

) i

& l

& '

 

where
&

2 31

 i

We know that is a martingale, so the integral for all . This implies that the integrand is zero; hence

must be zero

(FK)

&

&

) X

& (T

!   c  

 l0

 

 T

) i


j j

   9A 9 B

!  

!6



) i
j

qi

qi

) i

mip w  ! 

Recall the Independence Lemma: If then

!   c  

 l0

Dene

where

184

is a function to be specied later.

is a -eld,

is -measurable, and

is independent of ,

With geometric Brownian motion, for

 B

9 9

!6



where

-measurable

e

e
1 1

Wqiq Wqiq 2 31

! 3 

The independence lemma implies


 

! 3 



 '

  '  

 '

Now where We thus have

, we have

independent of

) i0 0 e 0
j j jg

i
j

  

i  e i'   e i'  e @ ' 

j j j

jn

e ' 


1

e q@
jn

2 31

!   c    e q@' 

) 1  )

! 1

 '

 

 T0

2 x1
CW

&

&

9A

) i

9 9 9 @

  9A 9  B 9

A9 9 9 9 2 71 B

BC 9A 9 9 9

!6    6 

!6

Note that the random variable whose conditional expectation is being computed does not depend on . Because of this, the tower property implies that , is a martingale: For ,

9A

2 g1

9 B C

9 @

!6  

This is a special case of Theorem 5.51.

This is a special case of Theorem 5.51 (Feynman-Kac).

) i
j

& l

& '

&

 2 2 2 31

 l

We have shown that

CHAPTER 16. Markov processes and the Kolmogorov equations

Because formula,

This leads us to the equation

Along with the above partial differential equation, we have the terminal condition

 '

 

 T

Furthermore, if condition

is a martingale, the sum of the

for some

&
j

at time if

Finally, we shall eventually see that the value at time of a contingent claim paying

. Therefore, , then also terms in . This gives us the boundary must be 0. By It s o is 185

2 x1
C

 

0
j

& '

     q l0  & 

jg

"e6

) & 0(

jn

B "

6

2 g1

 
j

In terms of the transition density

Even if is some other function (e.g., satises the Black-Scholes PDE (BS) derived above.

 7

 

p#

  !     e q 

qi

The Feynman-Kac Theorem now implies that

 cl

e q@
jn

"

 '

jn

and

In the case of a call,

for geometric Brownian motion (See Example 16.4), we have the stochastic representation

Compare this with the earlier derivation of the Black-Scholes PDE in Section 15.6.

Plugging these formulas into the partial differential equation for and cancelling the pearing in every term, we obtain the Black-Scholes partial differential equation:

186

16.7 Black-Scholes with price-dependent volatility

 e q@

also satises the terminal condition


c 0
j

0
j

&

, a put),

is still given by and (SR) (BS) ap-

6

T  l0
g

CHAPTER 16. Markov processes and the Kolmogorov equations

187

and the boundary condition

An example of such a process is the following from J.C. Cox, Notes on options pricing I: Constant elasticity of variance diffusions, Working Paper, Stanford University, 1975:

2 31

2 31

T0 1 '  i

where . The volatility sponding Black-Scholes equation is


c  2 1

decreases with increasing stock price. The corre-

& |

2 31

& 

 l 1

 c

(&

188

Chapter 17

Girsanovs theorem and the risk-neutral measure


(Please see Oksendal, 4th ed., pp 145151.) Theorem 0.52 (Girsanov, One-dimensional) Let , be a Brownian motion on a probability space . Let , be the accompanying ltration, and let , be a process adapted to this ltration. For , dene
2 2

and dene a new probability measure by


) r

Caveat: This theorem requires a technical condition on the size of . If


r

everything is OK. We make the following remarks:



is a matingale. In fact,

189

& 

& (

& (

& 

# qi

) i

& (

& (



2 1

Under

, the process

, is a Brownian motion.

& 

2 x1

2 1

&  

& 

2 1

#  

& 

mi 

#  

 |

&

2 31

190
1 1


so
 

is a probability measure.

in terms of

. Let

denote expectation under

. If

is a random variable, then


) "

Now use Williams standard machine. and


r
  

. The intuition behind the formula


&  
#

but since and , this doesnt really tell us anything useful about we consider subsets of , rather than individual elements of .

Distribution of

. If is constant, then
)i mi  

Removal of Drift from . The change of measure from To see this, we compute
C qi      

to

removes the drift from

(Substitute

&

mi

) & 

&

$ %

&

7mi`

mi

"mi` !

qi

"

   &

d#

Under , variance :

is normal with mean 0 and variance , so




is normal with mean

 

 C

 |

 

is that we want to have


1  |
  

) "

 

 &

 

#  

&  

 

 

To see this, consider rst the case

 #   

, where

. We have

 

3  

&    


 

 #  

is a probability measure. Since





, we have

for every

 

3 

 

 

 

 

 

 

 

 



. In particular

. Thus,

and

CHAPTER 17. Girsanovs theorem and the risk-neutral measure


1

191

We can also see that


  2


Because
&qi ! G% $ &qi ! $% &qi ! 
$ %

we have



Means change, variances dont. When we use the Girsanov Theorem to change the probability measure, means change but variances do not. Martingales may be destroyed or created. Volatilities, quadratic variations and cross variations are unaffected. Check:
)

17.1 Conditional expectations under


2 31

Lemma 1.53 Let

. If

is
  A

-measurable, then
) RR

Proof:
qC

2 31

because

, is a martingale under

& 

&) & (x

  " 

qC

( ) '

&

 

 C

& |

 

 

 

 

&

&

Under mean

, is normal with mean zero and variance and variance .

. Under

)  &

)  &

qi " )  &  qi "  7qiV  & 

qi

"

  &

  &

 

 

by arguing directly from the density formula

is normal with

) X A

 

 C 2

2 31

g 

   

&

&

& w & & r &

)g(& (& &

) i

& (&


& |

   

& & A

2 x1

 C

 

Although we have proved Lemmas 1.53 and 1.54, we have not proved Girsanovs Theorem. We will not prove it completely, but here is the beginning of the proof.

Lemma 1.55 Using the notation of Girsanovs Theorem, we have the martingale property

 &

     A  &  C

 

8C A 8C

       

    
 6

B C

 

  A

A ) RC 2

 C

 

 

 

Lemma 1.54 (Bayes Rule) If

192

is

-measurable and

2 31

Next we use Bayes Rule. For


& A  

Proof: It is clear that property. For , we have

Therefore,

Proof: We rst check that

is a martingale under

is

, (Lemma 1.53 again) -measurable. We check the partial averaging . Recall (Taking in what is known) (Lemma 1.53) , then (1.1)

CHAPTER 17. Girsanovs theorem and the risk-neutral measure

193

Denition 17.1 (Equivalent measures) Two measures on the same probability space which have the same measure-zero sets are said to be equivalent. The probability measures dened by and


of the Girsanov Theorem are equivalent. Recall that


)
5

17.2 Risk-neutral measure


2 2 w1

Stock price:

The processes and are adapted to the ltration. The stock price model is completely general, subject only to the condition that the paths of the process are continuous. Wealth of an agent, starting with several ways:

Risk premium
6

Market price of risk=

@ )9

& l

(&

S3

&

e

u3

& 

& 

78

(&

Capital gains from Stock


5

Interest earnings

& C

 x

k & (

& l

& l

& l

2 31

Interest rate:

. The process

is adapted. . We can write the wealth process differential in

As usual we are given the Brownian motion: , with ltration dened on a probability space . We can then dene the following.

) i

& 

2 a1

3 4

& 

1 2 

(&

S3

 

If

, then

 &

 |

 &

&

 

If of

, then to obtain

Because

for every , we can invert the denition

&  |

#  

 

& 

is

 W

 

Denition 17.2 (Risk-neutral measure) A risk-neutral measure (sometimes called a martingale measure) is any probability measure, equivalent to the market measure , which makes all discounted asset prices martingales.


eeT

ee

) i

' &

S3

 

&

&

 &

S3

) i

& 

#  |

) C

(&

&

u3

& 

5 5 Q

   

&

(& & l

& |

S3

& 

3 u4

(&

& l

&

&

&
j

& 

 

&

B F D D G E 

B I  F D D  R  |

) C

B D  R I F D

&

5 Q

& C

& l

B D  E H  F D B D  E H  F D

B I F D D P E 

&

&

&

& 

B D G E C F D

Under Then 194

Changing the measure. Dene

The discounted formulas are

Notation:

Discounted processes:

TS

and

are martingales.

CHAPTER 17. Girsanovs theorem and the risk-neutral measure


For the market model considered here,

195

Look back

This is because

is a martingale under
 1

, so
) B s 6
X

 

) B

6  

2 1

 

 l

If there is a hedging portfolio, i.e., a process , then satises


1

h v e D Q e G iHg(fl`

Q v Q G `G egH|8'aS

Q Q G ` |Ce|lbv

Q IG RP'`

# 3dS c

s th r p iqS

G S

 l

Example 17.1
Y

European call European put Asian call

, whose corresponding wealth process

Risk-neutral valuation. Consider a contingent claim paying an at time .

is the unique risk-neutral measure. Note that because .


1

we must assume that

-measurable random variable

W %

u3

& 

U  e me 3 e V

& (

qi| 

where

&  |

 # 

eeT

 '

196

Chapter 18

Martingale Representation Theorem


18.1 Martingale Representation Theorem
See Oksendal, 4th ed., Theorem 4.11, p.50.


then is a martingale. Now we see that if is a martingale adapted to the ltration generated by the Brownian motion , i.e, the Brownian motion is the only source of randomness in , then

18.2 A hedging application


2 2 1


Homework Problem 4.5. In the context of Girsanovs Theorem, suppse that the ltration generated by the Brownian motion (under ). Suppose that is a Then there is an adapted process , such that
A ) 2

is -martingale.

197

2 g1

l &

2 1

 

for some

& |

& (

 l

 '

Remark 18.1 We already know that if


&

&

In particular, the paths of

are continuous. is a process satisfying

2 31

A 2 2 1 2 31

Theorem 1.56 Let be a Brownian motion on the ltration generated by this Brownian motion. Let relative to this ltration. Then there is an adapted process

. Let , be , be a martingale (under ) , such that

2 31

& 

 l

2 31

) X

g3
5

 A9 9

 

 

 l

) 2

2 g1

l &

 

2 g1

) i

 &

&

A )

   |

& 

& |

 & # mi

& |

 

 |

& 

mi

 

& 

& 

1 2l 

3 4

&

Let

Then

198

be a portfolio process. The corresponding wealth process

According to Homework Problem 4.5, there is an adapted process

2 71

2 31

B R

9 9

6  
X

 

) X 2

 l

2 1
5

Let be an -measurable random variable, representing the payoff of a contingent claim at time . We want to choose and , so that

&

u3

&

5 u

& 2 1

Set i.e.,

Dene the

-martingale

and choose
5

vb e T

so that , such that satises

   |

h &

8 &

8 

# 

& ) "i

)  & # mi 

) ) F)

& 

#  

&

2 1

) ) F)
C

 | 1 2

A
y

&

) ) )

& 

& 

qi

A9 2
C

2 x1

9 9

B C

A9 9 9 9

   

9 X 6 w   6   X A 9 9 9 9 6
X

 

The Martingale Representation Theorem guarantees the existence of a hedging portfolio, although it does not tell us how to compute it. It also justies the risk-neutral pricing formula

) X A9

 '

9 9

 B

s 6
X

 

A9 )

2 31

9 9

6
X

 

 

 2

With this choice of

2 1

18.3

where

so

In particular,

CHAPTER 18. Martingale Representation Theorem

Theorem 3.57 ( -dimensional Girsanov) dimensional Brownian motion on

2 1

&

For

2 31

-dimensional Girsanov Theorem

dene the accompanying ltration, perhaps larger than the one generated by , we have ; , -dimensional adapted process. ,a 199 ;

200 Then, under




, the process

is a -dimensional Brownian motion.

Corollary 4.59 If we have a -dimensional adapted process then we can dene and as in Girsanovs Theorem. If , is a martingale under relative to , then there is a -dimensional adpated process such that

18.5 Multi-dimensional market model


, and let following:
A A 2 2 1
F

Let
)

, be a -dimensional Brownian motion on some , be the ltration generated by . Then we can dene the

Stocks

Accumulation factor
) h &  

C1

Here,

and

are adpated processes.

) ) F)

) ) F)

) ) F)

2 1

2 x1

2 31

& |

2 31

&

& ) (@i

@i & )

) ) F)

If , is a martingale (under -dimensional adpated process

) relative to , such that

the ltration generated by the Brownian motion

& |

 |

&

 qi

1 C 

 

 

&

2 1

) ) ) F

2 g1

C&

Theorem 4.58 on
21 A )

a -dimensional Brownian motion .

&

) ) F)

2 1

2 x1

18.4

-dimensional Martingale Representation Theorem

2 g1

) ) )

 

&

, then there is a

&

CHAPTER 18. Martingale Representation Theorem


Discounted stock prices

201

Risk Premium
3 C F

Case I: (Unique Solution). For Lebesgue-almost every and -almost every , (MPR) has a unique solution . Using in the -dimensional Girsanov Theorem, we dene a unique risk-neutral probability measure . Under , every discounted stock price is a martingale. Consequently, the discounted wealth process corresponding to any portfolio process is a martingale, and this implies that the market admits no arbitrage. Finally, the Martingale Representation Theorem can be used to show that every contingent claim can be hedged; the market is said to be complete. Case II: (No solution.) If (MPR) has no solution, then there is no risk-neutral probability measure and the market admits arbitrage. Case III: (Multiple solutions). If (MPR) has multiple solutions, then there are multiple risk-neutral probability measures. The market admits no arbitrage, but there are contingent claims which cannot be hedged; the market is said to be incomplete. Theorem 5.60 (Fundamental Theorem of Asset Pricing) Part I. (Harrison and Pliska, Martingales and Stochastic integrals in the theory of continuous trading, Stochastic Proc. and Applications 11 (1981), pp 215-260.): If a market has a risk-neutral probability measure, then it admits no arbitrage. Part II. (Harrison and Pliska, A stochastic calculus model of continuous trading: complete markets, Stochastic Proc. and Applications 15 (1983), pp 313-316): The risk-neutral measure is unique if and only if every contingent claim can be hedged.
 

Market price of risk. The market price of risk is an adapted process satisfying the system of equations (MPR) above. There are three cases to consider:
 

) ) F)

 x

) ) F)

For 5.1 to be satised, we need to choose

, so that (MPR)

& |

&

e & F

) ) F)

&

&

C1

1 C 

 

&
3

(5.1)

F t

202

4 1 )
3 1 4

& 

 mi

) &  & &  & & 3t'e 3 3 3 't &  3  & & & & & & 3 l 3 &  (& (&  & & 3 j 3 2 2 1 ) ) B &
3

Ql

3 d

(& &

3 T

& 1 6  & &


1



&

A 2 A

Let

In what follows, all processes can depend on and , but are adapted to simplify notation, we omit the arguments whenever there is no ambiguity.

2 1

2 31

A two-dimensional market model

Chapter 19

be a two-dimensional Brownian motion on be the ltration generated by .

2 g1

Stocks:

3 w

We assume

In other words,

has instantaneous variance

has instantaneous variance

3 3 e

and

have instantaneous covariance

Accumulation factor:

The market price of risk equations are

Note that

203

3 e

(MPR) . Let . To

g iS g hS

f S f #S g iS g hS f S f #S

F F F y y

A9 )

2 x1

9 9

6
X 

 

 |

A
X

) B & 3

&

6
5

& @

&

3 &
5

5 5

   &
5 

& @

(&

&

&

&

& 
5

&

&

We have changed the mean rates of return of the stock prices, but not the variances and covariances.

) B & 3

& &

3 e 3 js

& 7 6  (&  &  (&

) i

&

&  

 

 

A ) h

& 

& 

&

  &| #   qi  

3 e

@ 3 3 1


3

3
1

provided

The solution to these equations is

204

Suppose

Then

19.1 Hedging when

m rqp4l n o n m

Let


is the unique risk-neutral measure. Dene

be

-measurable. Dene the .

. Then (MPR) has a unique solution

-martingale ; we dene

CHAPTER 19. A two-dimensional market model


The Martingale Representation Corollary implies
) & 

205

We have
&
3 e

We solve the equations



3

Every
A

-measurable random variable can be hedged; the market is complete.

The stocks are perfectly correlated. The market price of risk equations are

1 

The process

is free. There are two cases:

&

&

3 T4

3 4

 1

&

& 1  &



&

The case

is analogous. Assume that

s uo

19.2 Hedging when




. Then

) X

 l

2 31

we have

and in particular,

 

 

 l

for the hedging portfolio

. With this choice of

and setting

& 3

# 

l
3

) &

&

3 l't

#
5

l 3

&

 |  &

&

 

(MPR)


X

) & & &


3 5

& @

2 1
1 2

3 g f g 3 U

'3

) q

 q

& B B & & 3 1 6

3 1 &

3 3
1 1

6 6

 

&

&


5

T3

5 g g 3 U f f 3 U

) q
5

 
g

&

& @

&

& @ &

3
5

&

& @

&

&
f

There is no solution to (MPR), and consequently, there is no risk-neutral Case I: measure. This market admits arbitrage. Indeed

) q

W

Case II:

206

have the solution

3 1

Let be an -measurable random variable. If be hedged. For example, if


&
5

(&

3 &

   &

w & @

3 '
1

5 5

&

3
1

Hedging:

Notice that

f U

and

Then

Suppose

is free; there are innitely many risk-neutral measures. Let

or

depend on

does not appear.

The market price of risk equations

Set

, then there is trouble.


Positive

depends on

be one of them.

, then it can probably not

 & T &

) &

&

&

& 

&

 |

A9 )

2 g1

9 9

B C

6  
X

 |

To get so

We can write

CHAPTER 19. A two-dimensional market model

More precisely, we dene the

to match

, we must have -martingale 207

208

&

&

qi

"

 

&

&

mi
{ | z

"
c

c }

&

&

h (&

mi

"

&

&

&

{ "

qi

z 

 
$

@! 

) i

@wxh  u v y
y

20.1 Reection principle for Brownian motion

With drift. Let

So the joint density is

Then we have:

Pricing Exotic Options

Chapter 20

Dene

Without drift.

209
s 

210

2m-b

shadow path

m b Brownian motion
Figure 20.1: Reection Principle for Brownian motion without drift

m m=b (B(T), M(T)) lies in here

b

t

Figure 20.2: Possible values of

) 

 G&

&

$ % $


 &

 &qi) !


 &
$

 
! 

 &

qi    mi !
$ %

 &

)
$  &

 z

&

 

 z

#
{

#
z

 '

!  

)
$  &

 z


v

&

 &qi3  !

 z

#
{


t

# 
z

$ %

eqiq !

  !  

  '

!  

) 

 &

&

qi

"



$

 &

&

t 

) i A

y i

w u  @~Q

) &    #    Cqi ! G$% Cqi ! $% Cqi ! 


$ %

where

2 31

Set

Let

Since


But also,

Under

CHAPTER 20. Pricing Exotic Options

is arbitrary, we conclude that be a function of two variables. Then is a Brownian motion (without drift), so , is a Brownian motion (without drift) on . Dene 211

(MPR)

Hh

k "e 3

t 3

qiq ! 3
! eqiq

T4  q 

eCf e

$ S 3 e e 

6 ' q   1    q T

) i

 D @w~| u  y i
y

G$

t 3

! &qi

 |

) i

T
3

 |



G$

&qi ! 3

A @ @ @ )9

k
3

7
7 3 6

78

qi  ! &qi 3

 

& 

& '

 @

3 w

e e

S  d

 ' qT  

) X

 w u @~} 
y y h S 

e e

Let

where

To simplify notation, assume that the option is

We compute,

(&


where,

Consequently, where Because


1

20.2 Up and out European call.

212

is the risk-neutral measure, be given. The payoff at time

is already the risk-neutral measure, so the value at time zero of is

& &

The standard method for all these integrals is to complete the square in the exponent and then recognize a cumulative normal distribution. We carry out the details for the rst integral and just


) T& (&

&

"

mi 

# 1

mi

 z

 z

m m m

" " "

#
 {

mi 
z  z {

# 1

&

&

mi


q
  {

"

"

qi

9 9 9  z

mi qi

! "mi` 3

 z

q  q  1

&

! qiq 3

#
z   {

qi

"

$ 3

! qiq

#
z

#
z 

 q w
z   { z

& & ) ) (3eF) 2

) 

We consider only the case

CHAPTER 20. Pricing Exotic Options

Figure 20.3: Possible values of

~ m

~ M(T)

~ b

(B(T), M(T)) lies in here

so

2 31

The other case,

leads to


2 g1

We compute

: and the analysis is similar.

x=y

x ~ B(T)

. 213

"e 3 

"e 3

 

h h 3 h





c

6
h


3 h

qi


3


c

 c

m 1

 w

)
h

 c

3 g

3 3


c



mi !
g

#
3  {

) X

& $ (&
3

& "

w3

mi 
z

#


1


mi

#
z

" q

&  &


w3 3 P

w3

 
3

3 u

03

where

Putting all four integrals together, we have to obtain In the rst integral we make the change of variable give the result for the other three. The exponent in the rst integrand is 214

CHAPTER 20. Pricing Exotic Options

215

v(t,L) = 0

+ v(T,x) = (x - K)

v(t,0) = 0

Figure 20.4: Initial and boundary conditions.


j fC

If we let

we obtain the classical Black-Scholes formula


4 tPeC d P qPuPeC

If we replace by and replace by in the formula for , we obtain a formula for , the value of the option at the time if . We have actually derived the formula under the assumption , but a similar albeit longer formula can also be derived for . We consider the function
e E

This function satises the terminal condition


I C

and the boundary conditions


eQPfC

We show that

satises the Black-Scholes equation


h d j

PEj

tedtI e# PEj eQ P e fC Pu 

HRe#  Qd

fC

$" " "

" e E

$" " " 1 " % &f % 4

1 % $" " "

te

de

5 76Q

t I

e

" e E

$tI "
" " 1 "

ea! f! Q

te3feb

(Ee

$" " "

" e E

t dtI

e#

Qdtefeb)fC

(Ee#

I' eb0% t






(Ee#Q

fCe#Q

tee R


$ t" H
" "

e I
$" " "

E


% f

" e E

te

e

I'f|h

I)f

(Re#v

% &

" e E

t dt

"Q

IfCQdt
$" " "

Ib'fh

IbfC

% &f

" #e E

e a!

 

hte "bfC

Re

dfe&iQx

Proof: First note that

Let

Let

is a martingale.

Theorem 2.61 The process

216

be given, and choose

be given and dene the stopping time

. If

 

On the other hand, if

e#f!

But when

  

Suppose

1 b2

In both cases, we have

so we may write

, we have

. Then , then the Markov property implies , then

QI Ct te

itff

udf

tPeCQP

f
V

C D8

Ct

 Qtf

e  8

hf

C H8

tff

fI  Q

A tff e 9 8

eQ

b'

1 T

"QIt

1 8

5  4e 

G F (

C H8

Q RP 5 I Be  1 8 (

G F

5  Be 

G 2 F ( E 7 Re

tPeQdeb'feb0% )

C D8

C 

e#uBt fCe# @8 A 9

Integrate from to

For

CHAPTER 20. Pricing Exotic Options

Because

teb'feb)f

E 7

(E

The Hedge

Let

then follows.

The PDE

is a martingale. Therefore,

be the wealth process corresponding to some portfolio

Then

and

We should take

, we compute the differential

is also a martingale, the Riemann integral

A stopped martingale is still a martingale

if if

. Then

. 217

W X U

eb

ueb

218
v(T, x)

v(t, x)

Figure 20.5: Practial issue.

20.3 A practical issue



Y

In particular, the hedging portfolio can become very negative near the knockout boundary. The hedger is in an unstable situation. He should take a large short position in the stock. If the stock does not cross the barrier , he covers this short position with funds from the money market, pays off the option, and is left with zero. If the stock moves across the barrier, he is now in a region of near zero. He should cover his short position with the money market. This is more expensive than before, because the stock price has risen, and consequently he is left with no money. However, the option has knocked out, so no money is needed to pay it off. Because a large short position is being taken, a small error in hedging can create a signicant effect. Here is a possible resolution. Rather than using the boundary condition
eQ fC tff Qf
V

solve the PDE with the boundary condition


eu
Y `

2. The value of the portfolio is always sufcient to cover a hedging error of size of the short position.

1.

remains bounded; times the dollar

f

where is a tolerance parameter, say %. At the boundary, short position. The new boundary condition guarantees:

t f

Qdf

f

For

but near ,

has the form shown in the bottom part of Fig. 20.5.

is the dollar size of the

e ve4

1
8

s %i

tf
C

e &cd

8 fu

d 8

G F

e 

8 f

G F

b c

f

C D8

fC

Dene the undiscounted expected payoff

QP

f 8

Payoff:

Introduce an auxiliary process

f

we have the solutions

With the initial conditions

Value of the payoff at time zero:

Stock:

Asian Options

Chapter 21

by specifying

219
it
d

b u # 

ut e

C
F

e cd

q p h f rig

1
8

G F

G F

b c

1
8

iP

bQ

31 8

G F

hf#Qf

8
V

U j 8

8
V C D8

U
U

C H8

8 4

H 8
V V


U 8

e Gb eHRe#QHGRef e

"Q

hcCt e b e h

e t

31 8

e v

iw

e ve

iw

e y"

The function

One can solve this equation rather than the equation for .

1 E

eHRe GR e

fC

G F

hf@

e b iGcGRef e

e b hcHGR e h 1

eQ

u ' 1

is the option value at time , where

x j 1 1
1

the terminal condition

Start with the stock price

21.2 Constructing the hedge

The PDE for

One can solve this equation. Then

and the boundary condition

21.1 Feynman-Kac Theorem

220

is

satises the PDE

. The differential of the value

of a portfolio

f

P

so that We want to have is (1.1)

"

1 8

thf 9 ht f

hPb eIEQPf

iP h

eCQI

e4

eQ

Pthe" e

e##  Q

Pthee

E 7

IEe u #

QCGR e

eC

8 f

b c

f I


u u

C H8

CH8 fbt f h 8 h 8 8 8

tffC j  8 8 Q
31 8

Compare this with

because both these processes satisfy the same stochastic differential equation, starting from the same initial condition.

C H8

fbf

8 f

U 8

G F

hfP 8

Take

The differential of the value of the option is

CHAPTER 21. Asian Options

21.3 Partial average payoff Asian option

1 8

G F

hf

Qf
U

If

PthPeQP

Now suppose the payoff is

tff

just as before. For

The hedge is given by and boundary condition with terminal condition


Qd

where

The function
x
V

at time . This value is

satises the Black-Scholes PDE . We compute , we compute next the value of a derivative security which pays off , then (From Eq. 1.1) 221

222 Remark 21.1 While no closed-form for the Asian option price is known, the Laplace transform (in the variable ) has been computed. See H. Geman and M. Yor, Bessel processes, Asian options, and perpetuities, Math. Finance 3 (1993), 349375.

)

Chapter 22

Summary of Arbitrage Pricing Theory


A simple European derivative security makes a random payment at a time xed in advance. The value at time of such a security is the amount of wealth needed at time in order to replicate the security by trading in the market. The hedging portfolio is a specication of how to do this trading.

22.1 Binomial model, Hedging Portfolio


8 

Evolution of the value of a portfolio:

so that



(four equations)

223

There are four unknowns:

. Solving the equations, we obtain:

d e

Given a simple European derivative security use a portfolio processes


V

, we want to start with a nonrandom

@V

 d

e4

Let Let

be the set of all possible sequences of coin-tosses. We have no probabilities at this point. be given. (See Fig. 2.1)

and

224

The probabilities of the stock price paths are irrelevant, because we have a hedge which works on every path. From a practical point of view, what matters is that the paths in the model include all the possibilities. We want to nd a description of the paths in the model. They all have the property

If we change , then we change , and the pricing and hedging formulas on the previous page will give different results. We reiterate that the probabilities are only introduced as an aid to understanding and computation. Recall: Dene Then

@V

i.e.,
x U

In continuous time, we will have the analogous equation


8

@V

x U

d

f

x U

@V

xU

x U

The paths of

accumulate quadratic variation at rate

per unit time.

PE

w R

PE

Let

. Then

p p qo r S E

k n

QRP 

% &

UI


U 8

PE

PE

PE

t d

R
1 E

k m l

QRP 

PE

UI  U

U U   It

8 8

8 U 8

j
1 1

It

PE
hi

fh g

d d
G G


V U

 


V U

CHAPTER 22. Summary of Arbitrage Pricing Theory


If we introduce a probability measure under which martingale, regardless of the portfolio used. Indeed,
@V x U

225 is a martingale, then

| 7{ U | 7{

We need to choose and so that


q
8 

The solution of these equations is


j
8

22.2 Setting up the continuous model


Now the stock price , is a continuous function of . We would like to hedge along every possible path of , but that is impossible. Using the binomial model as a guide, we choose and try to hedge along every path for which the quadratic variation of accumulates at rate per unit time. These are the paths with volatility . To generate these paths, we use Brownian motion, rather than coin-tossing. To introduce Brownian motion, we need a probability measure. However, the only thing about this probability measure which ultimately matters is the set of paths to which it assigns probability zero.
f PE f f hf 

j 1  } x 8  

x 8 

q%

%

$"

"

"

"

x !

x}

y

%

, and compute

| {

To nd the risk-neutral probability measure

~ l

under which

is a martingale, we denote

$"

"

"

"

y a

%$ "

"

"

"

xU

xU

xU G

xU

Suppose we want to have must have

, where

is some

-measurable random variable. Then we

$ "x "

"

"

G $ "w " Q t"P

"

will also be a

x U

a $" " "

"

U x

z y

226



, the paths of per unit time. We want to dene



C

so that the paths of accumulate quadratic variation at rate per unit time. Surprisingly, the choice of in this denition is irrelevant. Roughly, the reason for this is the following: Choose . Then, for ,
w

Since we are interested in hedging along every path, except possibly for a set of paths which has probability zero, the choice of is irrelevant.

The most convenient choice of so and

is
i

f

C D8

8 f

df 8

is a martingale under

. With this choice of ,

then this set of paths has positive probability when

p h if 7 Pudf

is dened by

f

If a set of stock price paths has a positive probability when



C

is dened by

f

In other words, regardless of whether we use or in the denition of paths. The mathematically precise statement is the following:

f

C g

p h 3if

p h if

PQdfe

~d

p h if

Pudf

is a continuous function of . If we replace However, there is an such that




by

, then

is a different function.

, we will see the same

 

f

G7

p h if

Pudf

PQdf

df

accumulate quadratic variation at rate

f

f

 

f
w

Let


, be a Brownian motion dened on a probability space

. For any

CHAPTER 22. Summary of Arbitrage Pricing Theory


z

227 is made, we have


C D8

and

is the risk-neutral measure. If a different choice of


f
S Q I RP 8 f df 8 p h if G PQdf C U

22.3 Risk-neutral pricing and hedging


z y

Let

denote the risk-neutral measure. Then


hf
C 8

Evolution of the value of a portfolio:


8 ftf
V

which is equivalent to
f f
x

Now suppose is a given -measurable random variable, the payoff of a simple European derivative security. We want to nd the portfolio process , and initial portfolio value so that . Because must be a martingale, we must have
i
V

This is the risk-neutral pricing formula. We have the following sequence:

% f

$"

"

"

"

E E Ea E {

Regardless of the portfolio used,



a

is a martingale under

f

C 8

f

f

f 8 f

8 f

tf

E { E

f

f

xU

xU

so

is a martingale under
E

C 8

f

Then
x 8

hf

where

is a Brownian motion under

. Set

has the same paths as . We can change to the risk-neutral measure Brownian motion, and then proceed as if had been chosen to be equal to

, under which .

fI
C H8 E

Q tP

fC

8
I

8 f

f 8

C C

E E {

is a

(3.1)

(3.2)

(3.3)

228 1.
a

is given,

V

under . We next use the corollary to the Martingale Representation Theorem (Homework Problem 4.5) to show that
f
C 8

for some proecss . Comparing (3.4), which we know, and (3.2), which we want, we decide to dene
fBf f
x

Remark 22.1 Although we have taken and to be constant, the risk-neutral pricing formula is still valid when and are processes adapted to the ltration generated by . If they depend on either or on , they are adapted to the ltration generated by . The validity of the risk-neutral pricing formula means: 1. If you start with
a C C a

Remark 22.2 In general, when there are multiple assets and/or multiple Brownian motions, the risk-neutral pricing formula is valid provided there is a unique risk-neutral measure. A probability measure is said to be risk-neutral provided

% f

$"

"

"

"

xU

2. At each time , the value

of the hedging portfolio in 1 satises

hf

then there is a hedging portfolio

, such that

and it will end with value


y x !

$"

"

"

"

From (3.3), the denition of

, we see that the hedging portfolio must begin with value

hf

Then (3.4) implies (3.2), which implies (3.1), which implies that the portfolio process .
hf
V

To carry out step 3, we rst use the tower property to show that
z y

dened by (3.3) is a martingale

hf
U

3. Construct so that (3.2) (or equivalently, (3.1)) is satised by the dened in step 2.
R { R f
V

f~

df

xU

hf

2. Dene

, by (3.3) (not by (3.1) or (3.2), because we do not yet have

). ,

(3.4)

(3.5) , is the value of

CHAPTER 22. Summary of Arbitrage Pricing Theory


it has the same probability-zero sets as the original measure;

229

it makes all the discounted asset prices be martingales.

To see if the risk-neutral measure is unique, compute the differential of all discounted asset prices and check if there is more than one way to dene so that all these differentials have only terms.
C 8 C

22.4 Implementation of risk-neutral pricing and hedging


To get a computable result from the general risk-neutral pricing formula
%
a x

one uses the Markov property. We need to identify some state variables, the stock price and possibly other variables, so that
% " "

is a function of these variables.

where
0D( F 0y

&y' ( )H

7 g

7B'Bt6

0H

046

46

Take

and

to be the state variables. Dene

D(

Example 22.2 Assume and

44t

and

is a martingale under are constant.

"

" y'

"

"

46Bt6

04

Then

y y'

y'``

( H6

0HBt6

Example 22.1 Assume and variable. Dene

are constant, and

$"

"

$"

" " "

y x

f f

xU U ( (

. We can take the stock price to be the state

C 8

G F

P

xU

C 8

In every case, we get an expression involving to be a martingale. We take the differential and set the term to zero. This gives us a partial differential equation for , and this equation must hold wherever the state processes can be. The term in the differential of the equation is the differential of a martingale, and since the martingale is

&4646Bt6

)0D(&'

4 46

" " " F " y'0&H&' " " % 4 " &46464t W

46

" y y'

046

hi h h fh 07Bx4t F I hh W

r S y' p p p p p p qo

6 Q RP )0D(&'

&y(x 4 46 04t'0&46&t646#

46&46&t6)446t4

4@464t " " " % 46

46 4

46@&44t " &y(x D

46

Then

Take

Example 22.3 (Homework problem 4.2)

is a martingale under

and

230

and

46

46

Then

is a martingale under

and

to be the state variables. Dene

we can solve for


8

f

. This is the argument which uses (3.4) to obtain (3.5).


.

46464t

4604&t6) &46&t6'

'&46&4Bt6

)4

46 4&t6)4 4 46&46Bt6

4 46 46 4

D46

7B'Bt6

D3

% 0 t % HD R

3(

##

t #

 R

4 46 46

0&46@&46Bt6

4&t6

( S 46&46Bt6

Q RP F I 46 46

0&Hx(

where it should be noted that

The partial differential equation satised by

is a martingale under

Example 22.4 (Continuation of Example 22.3)

CHAPTER 22. Summary of Arbitrage Pricing Theory

. We have

, and all other variables are functions of

4t

where (see (3.2))

is

, and

. We want to choose

4`

4&t6'

Therefore, we should take


&46&t6)`

to be . We have so that

231

232

Chapter 23

Recognizing a Brownian Motion


2. 3. Then

is a martingale,
f f

, (i.e., informally

is a Brownian motion. is normal, with . We shall show that the

Since the moment generating function characterizes the distribution, this shows that is normal with mean 0 and variance , and conditioning on does not affect this, i.e., is independent of . We compute (this uses the continuity condition (1) of the theorem)
f
C H8 E

so
S Q P RtRI 8 5

uses cond. 3

233

f

if

Proof: (Idea) Let be given. We need to show that mean zero and variance , and is independent of conditional moment generating function of is
$

C H8

1 d

'E



5 u

C H8

5 i

C H8

1 d f

$"

" E E

"

"

C H8

F 2 E

C H8

5 i

5 #i

5 Q

5 8

f

1. the paths of
C

are continuous,

).

Theorem 0.62 (Levy) Let , such that:


C

be a process on

, adapted to a ltration

d C

if
C

$" " " " (E E " " " E 5

5 Q

$"

5 #

E 'R

5 Qd

5 x

5 ec

5 ecf
1

hf

f

| 8 

5 Qf

5 Qd

$" " " " E 6 5 if QdG

% &

$" " " " E 6 5 a! F

$" " " " E 1

% &

5 Q

% &

5 f

$" " " " % 4

G
E 5 i 1 G F !

Q q

C D8

$ " 
" " "

C D8

5 i

1 5 1

G F F

% 4



C H8

C H8

Now



We dene

so that

Plugging in , we get

Therefore, and It follows that 234

is a martingale (by condition 2), and so

8
C D8

t  t
C D8

C D8


C D8

8 8

8 8

8 8


C D8

C H8

C D8

C D8

t t
C D8

8  

C H8


C H8


C D8

8 8


C D8

C H8

t t

C H8

C D8

t
C D8

8 8
C C

C H8

Dene processes

23.1 Identifying volatility and correlation

CHAPTER 23. Recognizing a Brownian Motion

Let

and

Dene

be independent Brownian motions and

and

The Brownian motions

Then

and

Therefore,

and similarly

and

have continuous paths, are martingales, and

are Brownian motions. The stock prices have the representation

and

by

are correlated. Indeed, 235

 8
A 8

8 9

8 9

C D8

C D8 C D8

8
8

C D8

~ C D8 ~

t ic

C D8

C D8


8 C D8

C D8

8 8

e~
C D8

e~ u

eQ

t t t t
n

8 8

8 8

Suppose we are given that

where

23.2 Reversing the process

236

and

are Brownian motions with correlation coefcient . We want to nd

If

This corresponds to

A simple (but not unique) solution is (see Chapter 19)

so that

, then there is no

and

Continuing in the case


t ic

t ic

, we have

CHAPTER 23. Recognizing a Brownian Motion

237

so both

and

We can now apply an Extension of Levys Theorem that says that Brownian motions with zero are independent Brownians. cross-variation are independent, to conclude that

Qd 8

C D8

C H8

are Brownian motions. Furthermore,

238

Chapter 24

An outside barrier option


Barrier process:
hf
d 8

Stock process:
C H8

at time , where

d

Remark 24.1 The option payoff depends on both the and processes. In order to hedge it, we will need the money market and two other assets, which we take to be and . The risk-neutral measure must make the discounted value of every traded asset be a martingale, which in this case means the discounted and processes.

239

C H8

C 8

C D8

C 8

We want to nd

and

and dene

ee

f

dB

iG }d

P4x

. The option pays off:




b2i%e

where

, and

and

are independent Brownian motions on some

C H8

C D8

f f
d

d ~

8 f
$

240 so that
C 8 E ee R

We must have (0.1)



We solve to get
x

We shall see that the formulas for and do not matter. What matters is that (0.1) and (0.2) uniquely determine and . This implies the existence and uniqueness of the risk-neutral measure. We dene

The value of the option at time zero is


dB H  e
y

We need to work out a density which permits us to compute the right-hand side.

i.e., the correlation between

and

is .

 8

e~

tP

hPe

Remark 24.2 Under both

and

has volatility

has volatility

and

Under , and are independent Brownian motions (Girsanovs Theorem). risk-neutral measure.

z y

C 8

C D8

C D8

8
d C 8

z 8

C D8

8 8 8 8

q if p h

(0.2)

is the unique

p h 3if

PQ

q
C

t


C 8

8 8

W 

p h 3if

8 d
d

g g

p h if p h if

P P

f if C

iG }d


Set


s R

df
C

gif q p h

P


C 8

d d 8

Then

From the above paragraph we have

p h 3if p h 3if

PQ PQ

C 8


d d

C iz

The joint density of

so so The stock process. so Recall that the barrier process is and

CHAPTER 24. An outside barrier option

, appearing in Chapter 20, is 241

Furthermore, the pair of random variables is independent of are independent under . Therefore, the joint density of the random vector is

ee
%

 E

E   k 
G

p h if

P
y

e ee E R

B H

We know the joint density of

24.1 Computing the option value

242

. The density of

is

because

C h

The answer depends on and . It also depends on depend on nor . The parameter appearing in the answer is

Remark 24.3 If we had not regarded as a traded asset, then we would not have tried to set its mean return equal to . We would have had only one equation (see Eqs (0.1),(0.2))

to determine and . The nonuniqueness of the solution alerts us that some options cannot be hedged. Indeed, any option whose payoff depends on cannot be hedged when we are allowed to trade only in the stock.

y C z

p h if

The option value at time zero is

and

. It does not

(1.1)

k

P

8 l y 8 8

p h 3if


W p h 3if

4 4 5 5

p h 3if

P

G F

e#Q heC

0 ' $ 2 3 1) (& #"!k k %


tP
d

C 8

and

C iz

e Q 
y

P

hPeC

CHAPTER 24. An outside barrier option


d

243 , then
h R
8 8 d

If we have an option whose payoff depends only on original equation for ,

is superuous. Returning to the

we should set
z

Now we have only Brownian motion, there will be only one , namely,

and we are on our way.

24.2 The PDE for the outside barrier option


Returning to the case of the option with payoff
ee
E

we obtain a formula for


e%
E 5 6

u u

u i

C 8

C 8

d 4

tf

if e

C 8

is a martingale under
8

. We compute

tf

if e#

P

and

t

iPeC

P

by replacing , and Now start at time 0 at process


P

by

, and respectively in the formula for . . Using the Markov property, we can show that the stochastic

E A C A@ FDB

8 6 97

Bt

u # 

eIEQGR e

y E

fC

9 He#Q

so with

we have

so

is a Brownian motion under

(Levys theorem), and

C D8

C D8

C H8

C D8

244

y L v(t, x, L) = 0, x >= 0

x v(t, 0, 0) = 0
"
  

Setting the

term equal to 0, we obtain the PDE

The terminal condition is

and the boundary conditions are

e4

Figure 24.1: Boundary conditions for barrier option. Note that

is xed.

~e

eQ

de

u u

"QIPef

u i

Ge

eQH

CGR e

f

e

8 
V

U  d

8 U V

U 8

Let denote the number of shares of stock held at time , and let denote the number of shares of the barrier process . The value of the portfolio has the differential
f
V

d 

 8

e

 8

tf

if e
8


 d 8

f
C 8

f
d

e u


8 f

e u e#u

e
8

% f C 8

f

C 8

f 8 f


f 8

e u

f e
8

d C
d


% u C 8

C 8

C 8

e Q

hffC e E
8

eQ

%

eHRe QHPefC

eQCPefCeQI

ue

CHAPTER 24. An outside barrier option

This is the usual Black-Scholes formula in .

This is the usual Black-Scholes formula in .

u u

the terminal condition is

The boundary conditions are

the terminal condition is

xe

eQ

Cte e

e

On the is

boundary, the option value

On the boundary, the barrier is irrelevant, and the option value is given by the usual Black-Scholes formula for a European call.

be

eQIt" e q

efC

where that

After setting the

24.3 The hedge

term to 0, we have the equation

The boundary condition is

, and

are functions of . Note

tf
d

h f

tf

hff

Therefore,
C
e

245

f
d

hff hff

Qdf Qdf
V

htf

tP

hPeQP

U d U

8 f
V

fe#

 8

To get

hffudf

and This is equivalent to 246 for all , we must have

Chapter 25

American Options
This and the following chapters form part of the course Stochastic Differential Equations for Finance II.

25.1 Preview of perpetual American put


C D8

25.2 First passage times for Brownian motion: rst method


(Based on the reection principle)

z C

Let

be a Brownian motion under

, let

be given, and dene


df

is called the rst passage time to . We compute the distribution of . 247


The plan is to comute and then maximize over to know the distribution of .

to nd the optimal exercise price. We need

if if

tt

I % Pe

fe4

C e4

iQi

I % 7i Qd

V

iQ

"

 ` 

Let

be given. Suppose we exercise the rst time the stock price is

tf

|

Intrinsic value at time




G H

or lower. We dene

f

S
8

p h 3if

U V 4 3T
F

" "

p h 3if

4

F

4 5 w R "" 3 wR


p h if

p h 3if

4
F

8 8

4 5 3

8

p h 3if

5 iG @

Dene

We make the change of variable

8

f

C 8

Now Therefore, From the rst section of Chapter 20 we have 248 Figure 25.1: Intrinsic value of perpetual American put

in the integral to get

K
Intrinsic value

Stock price

ji

p h 3if

gz u d
z y

iz
y

hf C

z 8

f C "

CiQx

C C

G G
C

p h 3if

f df C

p h 3if

e Q

gz

X `

G F

I X Y5

p h 3if

p h 3if



F v

S
8

p h 3if

U V 4

W
z gz

W W W

f 8

gz

so

For

Reference: Karatzas/Shreve, Brownian motion and Stochastic Calculus, pp 196197.

25.3 Drift adjustment

Reference: Karatzas and Shreve, Brownian Motion and Stochastic Calculus, pp 95-96.

Under

We x a nite time xed, dene

Dene

We also have the Laplace transform formula

CHAPTER 25. American Options

, the process

, dene

and change the probability measure only up to

, is a (nondrifted) Brownian motion, so


$

(See Homework)

. More specically, with 249

f
C

Qx

e4

Q 8

p h if

G F

I X Y5

df

C e&iQx

ji

p h 3if


8 8

#
8

p h if

G F


G F

#
W

p h if


#

p h 3if G F

y #

u
y y y y y


% % b

p h if p h if

$ "
" "

d d

"

p h 3if

p h if

I be a I b a I ab| I a b

I ab I a be

For

ji

Since

Therefore,

250

is arbitrary, this must in fact be the correct formula for all

For for nondrifted Brownian motion:

25.4 Drift-adjusted Laplace transform


we have .

Recall the Laplace transform formula for

C g

p h 3if

df

c V
q

b z

f f c Dc etViQ c T T

e hQ

c V

c !X

et iQ%

c V

I 5#i X e YX
G

4 I a

#

#

I 5 X

e YX

Y

Pq

#

I X Y5

Y

"

'

d
Y

the Laplace transform is

where in the last step we have used the formula for

CHAPTER 25. American Options

with

replaced by

c dPX

T 4

e Q
G F

c V

p h 3if

c V
G F

p h 3if

p h 3if

G F

I 5 X

If

, then

I 5# X

e YX

I ` X

if

, then

for every

#

Therefore,

I ` X

(Recall that

Letting

e
Y

If

If

we obtain

25.5 First passage times: Second method

Let

(Based on martingales) , then , then and using the Monotone Convergence Theorem in the Laplace transform formula ). , so . 251

be given. Then

ggg p

c
f
C p

ih

Q RP

gggi h

I fhh

ggg

g gg

p h f 3in

p qo

p h f 3in

df P 8

C D8

e4

e4

eb0f

p h 3if

e


ef

eh

eb

ef

e)f

for the rst passage time for nondrifted Brownian motion.


Y

I X Y5#i QiX

# #

i Q

4 I

p h 3if

Q Q

deb0f

C #

ebf

deb0f
p h if

C g d


p h if

de0f


C #

eb

C g

p h 3if

deb0f

C #

p h 3if C

eb C g de)f

eb)f

ebg

There are two possibilities. For those

this is justied by the Bounded Convergence Theorem. Therefore,

We want to take the limit inside the expectation. Since

is a martingale, so

252

is also a martingale. We have

for which

p h 3if

p h if d

Q
d

eb0f
d

where we understand

Therefore,

For those

for which

to be zero if

Let

25.6 Perpetual American put


T
e

, so

. We have again derived the Laplace transform formula , . ,

U 

y w x

u F u Y

Intrinsic value of the put at time :

We want to choose the largest possible constant. The constant is


'

U 

v 9

U 

w x

v 9
h

V V

s t

s t


p h 3if

c dt

PE

q r

PE

V V

I % 7e

i t

PE PE

f
C

"

& & &

Q Q Q

"

df"

tfV

Let

"

The curves Therefore,


  T

We compute the exponent

Dene

CHAPTER 25. American Options

be given. Dene for

are all of the form

. 253

254

value

K-x K 2 (K - L) (x/L)-2r/

Stock price

Figure 25.2: Value of perpetual American put

value

C3 x C2 x C1 x

-2r/ 2

-2r/ 2 -2r/ 2

Stock price

Figure 25.3: Curves.

U 

'

'

v 9

ec %

V

V

W W

and

Since

We have Note that


x

where to get We solve

Solution to the perpetual American put pricing problem (see Fig. 25.4):

CHAPTER 25. American Options

we have 255

f gYq
(7.4)

In other words, solves the linear complementarity problem: (See Fig. 25.5).

k ~x~p d q

 t  pd q qo o k q p k 51o n

h f ~ p x k o d q t q q x ~ q q dk q ~xk k x ~ d k ~ q q o k p1o k 1d k n p k n q p51o1#n k p5k11k 51dn k o n p ko q


, then , then

p d xk 5k o p o q q r q

256

25.7 Value of the perpetual American put

Set

b p o (h f p h HVke 1o D`Y p o 7(~ qr j e } { x ~dk | p w Vp wz x o  o y 1dn p ko qv t ulskQi k p ko 51dn q r de 5 f g 9 Y


. If , then

If

mQjh l k i

If

mQjh l k i

If

t l k Qi

where , then

Figure 25.4: Solution to perpetual American put.

L*

K-x

Stock price

-2r/ (K - L* )(x/L* )

(7.3) (7.2) (7.1)

value

CHAPTER 25. American Options

257

Figure 25.5: Linear complementarity For all ,

One of the inequalities (a) or (b) is an equality.

and is the boundary between them. If the stock price is in , the owner of the put should not exercise (should continue). If the stock price is in or at , the owner of the put should exercise (should stop).

p o

p o

or equivalently,

p s o

p o

p p o p os p o o q q p Q o pp h o o dn

Let be given. Sell the put at time zero for stock and consuming at rate at time . The value

. Invest the money, holding shares of of this portfolio is governed by

o p o p 7(~ o p s P(~ p o 7(~ q

25.8 Hedging the put


p h 1o

hQ n k k q n q n k p5k owp51odnk k qr

p o

p os p Q o

p o

p ut h

The half-line

is divided into two regions:

pk o n qv h n k 5k q n q n

gk

sk


(a) (b) (c)

258

We should set

To hedge the put when , short one share of stock and hold in the money market. As long as the owner does not exercise, you can consume the interest from the money market position, i.e.,

3.

is the smallest process with properties 1 and 2. be a supermartingale satisfying

We use (8.1) to prove (8.2) for

, i.e.,

pp h o 1o !n

pp h 1o

z p ho o |{Dy 1 q

w1o i p h

Case I:

We have

Proof of (8.3), assuming

is a supermartingale satisfying (8.1):

p o

If is not zero, we can take to be the initial time and adapt the argument below to prove property (8.2).

to be the initial stock price, and then

t l

p h o p1o !n 1s p ho f h ijh p o o!n7(~ p p o Q

Then property 3 says that

t Ql

i h

p o p

o 7(~ p s o qr Q

Explanation of property 3. Let

t Ql

i jh

2.

1.

p p o !n 7(~ o p o p o P(~ p p o !n 7(~ o qv p Q p o o !n 7(~ p p o dn 7(~ o

Properties of

is a supermartingale (see its differential above). ;

f d q

p p o 1n o p m o



p o

p o

lp o p p o !n o qr l Hp o

Remark 25.1 If

, then


(8.1) (8.2) (8.3)

P5( p o p p o 1p o o n  77(~ P(~ f Yq p o p o o p o 7(~ p n  n o p p o o p o n p 7p o dn 7~ o p p p o o p o p o !n q 7(~ p p o n p s o o f (  po Y f

The discounted value of the put satises

CHAPTER 25. American Options

259

25.9 Perpetual American contingent claim


p p o x o
Intinsic value:

where the supremum is over all stopping times. Optimal exercise rule: Any stopping time Characterization of :

which attains the supremum.

25.10 Perpetual American call


q p o 57~ o (

Theorem 10.63

k !

p ko 51dn

p ko 51dn

o !n 7(~

3.

is the smallest process with properties 1 and 2.

t Ql

l jh

p o P(~ p o

o !n 7(~

2.

pp o pp o p p o

o !n 7(~

1.

is a supermartingale; ;

p o x 7~ p o

Value of the American contingent claim:

pp h o 1o !n f 5Y p p | {z y o r o 7(~ q P Q 5Y o Dp

(Fatous Lemma) (by 8.1)

(See eq. 7.2)

5Y

(D

p o D

p ko 51dn

Q 1s p ho

Now let

to get

(Since

5Y

o Dp p h o pt o Q 1s Q 1o p h

h u

Case II:

: For

, we have (Stopped supermartingale is a supermartingale) )

D z | Y(1 y { (

p o p o q r e h
.

p ko 1dn

ki7p o

7~

l Q p

p o 57~ o

p o 7(~ p o e
and dene

k po h k pk1!n o

7~ k f q g 7(~ p o 7(~ q p p o o 7(~ e q p ko p o o 7(~ Q 51dn p q j Q

Now start with

Let

Then:

Proof: For every ,

260

1.

is a supermartingale (in fact,

p1o !n 1 p h o p ho p p o o 7~ p o q Q

2.

t Ql

i jh

Therefore,

to get

, i.e.,

p ko 1!n
.

Value of the put:

~ p p o r o (Y ~ f q k p o

See Fig. 25.6. It can be shown that jump.

There is no optimal exercise time.

Remark 25.2 No matter what

Intrinsic value:

Expiration time:

25.11 Put with expiration

we choose,

(value of the put at time if

stopping time

is a martingale);

are continuous across the boundary, while )

n n n

Y 5k dn p o

Let

p h 1o

be given. Then has a

! h

p k o

p1ox k n h n k n k n n q q q

p p o x1Y (~ o ~

Y(1 (

p o 5k dn
Value of the contingent claim:

pp o x o h e

i k i sQjh k

q r

p k !n o

n k n k n n Yq q h n n h n d q k q r n

k h p k !n o n k n h n k n q k n q r k

CHAPTER 25. American Options

1.

q ui

p p o !n 7(~ o o 7(~ p p o !n 7(~ o Q i h p o !n 7(~ p o

2.

ui

i jh

p o p

Then

Intrinsic value:

25.12 American contingent claim with expiration

Expiration time:

3.

At every point . .

p ut h

Characterization of : Let

p h 1o

is the smallest process with properties 1 and 2.

Figure 25.6: Value of put with expiration

be given. Then is a supermartingale; , either (a) or (b) is an equality. ; 261 (b) (c) (a)

2. 3.

If

, then there is no optimal exercise time along the particular path .

p o x p o

p p o

The optimal exercise time is

d(h o n

bdje

p p o !n 7(~ o p p o 7(~ o p p o o !n 7(~ Q i h p o !n 7(~ p o

is the smallest process with properties 1 and 2.

ui

1.

262

p Bo

is a supermartingale; ;

Chapter 26

Options on dividend-paying stocks


26.1 American option with convex payoff function
p

where and are processes and a.s. This stock pays no dividends. be a convex function of , and assume . (E.g., ). An Let American contingent claim paying if exercised at time does not need to be exercised before expiration, i.e., waiting until expiration to decide whether to exercise entails no loss of value.

263

p o

pp o o

pp o o x

p o

Consider a European contingent claim paying is

p o p o i po ppo x o p o p o f

at time

p o

p o

p to

and

, so

. The value of this claim at time

d i

i h

p 5o

sq

po i jh p o

Let

be the time of expiration of the contingent claim. For

p ko 51x f p51ox p h o p k o 1x q d i p k x x d x 51x hp oo p k o q

Proof: For

and

, we have

k 1o

p ko 1x

p o h
(*)

pp o o p ho 1 k h p o i ih h o p o Q p o p o p o

Theorem 1.64 Consider the stock price process

i i d Qjh p ko 51

264

Figure 26.1: Convex payoff function Therefore,

(by (*)) (Jensens inequality)

is a martingale)

This shows that the value of the European contingent claim dominates the intrinsic value of the American claim. In fact, except in degenerate cases, the inequality

is strict, i.e., the American claim should not be exercised prior to expiration.

26.2 Dividend paying stock

Let and

be constant, let be a dividend coefcient satisfying

pp ko k 1x 1o

ui

p o p o p  p o o p o p p o po o p o pp o o

i jh

p o

l f d

p o ox p Q p o

l h

p p o f p o p o po p o

p Q o p o o x d p o d p o d p o d p o p o p Q o

p ko 51x p k o 51x p p o x o

.... .... . ..... ......... .. ..... . ..... ....... . . ..... . .... .... .... .... .... .... .... .... .... .... .... .... .... ..... ..... ..... ..... ..... ..... .. . . ...... . . . . . . . . . . . ..... . . . .. .. .. . . . .... .... .... .... .... .... .... .... .... .... .... .... .... ..... ..... ..... ..... ..... ........ . . .... ...... ..... . . ...... . . . . ....... . . . . ........ . . .......... . . . . . . ........... . . ............ . . ..............

CHAPTER 26. Options on dividend paying stocks

265 be the time of dividend payment. The stock

Consider an American call on this stock. At times , it is not optimal to exercise, so the value of the call is given by the usual Black-Scholes formula

This function satises the usual Black-Scholes equation


%

and boundary condition

The hedging portfolio is

p h o

1.

p p o

Proof: We only need to show that an American contingent claim with payoff need not be exercised before time . According to Theorem 1.64, it sufces to prove

k 85Y 0 p k p

p o o n ui l p p m o f i jh p o o i p % ui jh h h o % i p f o dn p o k 1o p 5k o q d q

( 7 6 &

Y e

(where

) with terminal condition

i jh

h k %

% 5

% q

i h

Theorem 2.65 For

, the value of the American call is

p p o o (73 ~ ~4 2 3 p k o % f e p p o o % 01kpY o dn p o k 1o p 5k q d q ( & )Y'e

where
%

p o p

p 5k o

At time

, immediately before payment of the dividend, the value of the call is

p p o Y p

o !n o q d

At time

, immediately after payment of the dividend, the value of the call is

, where

p # $

! "

o Vp

o k

where

at time

ui

l i jh

p t o pp o p o Y d o p p o o p o Vp q o q u q q p o p o p ( o p h o u q ( 5k pp q o ~ u~ o  (~  p k o q d   q  o uk o  q pp 5k q p o k !n

p ho 1e

be an expiration time, and let Let price is given by

266

Since

, we have immediately that

To prove that is convex in , we need to show that is convex is . Obviously, is convex in , and the maximum of two convex functions is convex. The proof of the convexity of in is left as a homework problem.

is the number of shares of stock held by the hedge immediately after payment of the dividend. The post-dividend position can be achieved by reinvesting in stock the dividends received on the stock held in the hedge. Indeed,

# of shares held when dividend is paid

dividends received price per share when dividend is reinvested

Case II: . The owner of the option should exercise before the dividend payment at time and receive . The hedge has been constructed so the seller of the option has before the dividend payment at time . If the option is not exercised, its value drops from to , and the seller of the option can pocket the difference and continue the hedge.

k 1o

pkpY o dn o k q d kq q

p s o

4 A 3 B

p s o

p o m

where

p pkpY o o p n o p 5k ospY q d o q d q d p k 5pY o dn 5k o p q d

Case I: . The option need not be exercised at time have

p o Y p po p o q d p s

o o %

k 1olHp5kpY o !n o q d

k o

o p s o q d o sY p s q d o

pkpY o !n o q d p o

p m o

p o s

Let

@ 9

26.3 Hedging at time


(should not be exercised if the inequality is strict). We

p5kpY d o !n o q 0 phpY h o dn p o h 1o q d q

& Y'e

p h o

o p5kpY d o !n k q

p 5k o

h h !n p o p 5k o %
%

2.

is convex in .

uq

k 1o

Chapter 27

Bonds, forward contracts and futures


F p G

Here, and are adapted processes, and we have already switched to the risk-neutral measure, which we call . Assume that every martingale under can be represented as an integral with respect to . Dene the accumulation factor
I ) H

p o

Given

dollars at time , one can construct a portfolio of investment in the stock and money 267

U V

p o

A zero-coupon bond, maturing at time , pays 1 at time to the risk-neutral pricing formula, its value at time

and nothing before time . According is

3o E

Let be a Brownian motion (Wiener process) on some sider an asset, which we call a stock, whose price satises

p o

Q R

p o p o Q p o p o p o

p o

p 53o

p3o P

q S T e p p o to p o e p o

p o (

p o

p o

ei

i h mp o
H

p o
C

C D

. Con-

268

martingale

If is random for all , then is not zero. One generally has three different instruments: the stock, the money market, and the zero coupon bond. Any two of them are sufcient for hedging, and the two which are most convenient can depend on the instrument being hedged.

p t o

i.e., . Then given above is zero. If, at time , you are given invest only in the money market, then at time you will have

f d

p 53o

p o p o p o p o q (

p 3o

p o

p o

p o

If

is nonrandom for all , then

i umi

f d 3s p o

p o

p o

p o

If, at any time ,

and we use the portfolio

p s o

We set

o p o p o p s o p Q7p o o o p C om p oQ p o p o p s p o o q

The value of a portfolio satises

, then we will have

dollars and you always

p o

p o p o

p o p o p o p o p o f pC P p h p3o 53o o tp o p o C p5o p3o P p h 1o p o C p p3o o p53o P p o d

p o p o f p o p o

p o

market so that the portfolio value at time

is 1 almost surely. Indeed, for some process ,

{ p z o

p o

p Q o

p t o

p o

(*)

CHAPTER 27. Bonds, forward contracts and futures

269

27.1 Forward contracts


We continue with the set-up for zero-coupon bonds. The -forward price of the stock at time is the -measurable price, agreed upon at time , for purchase of a share of stock at time , chosen so the forward contract has value zero at time . In other words,

This implies that


W

27.2 Hedging a forward contract


Enter a forward contract at time 0, i.e., agree to pay for a share of stock at time . At time zero, this contract has value 0. At later times, however, it does not. In fact, its value at time is

p h o

This suggests the following hedge of a short position in the forward contract. At time 0, short -maturity zero-coupon bonds. This generates income

p o

p o W p o

Remark 27.1 (Value vs. Forward price) The -forward price the forward contract. The value of the contract at time is zero. time which will be paid for the stock at time .

is not the value at time of is the price agreed upon at

p o

po p h 1o p o p o 1o pp h

p h 1o

po p o p o

P P 

p h o

p o p o p p o

p h o

p f p o p o W aq p o

W X

p o p o p o f q W p p o o p to q W p o p o o

p h 1o p h 1o

W X

o 1o p h W Xq p o ph1o W Xq p o p to po p o

We solve for

ui

i h

p o

p p o

W X

p o p o f p o p o p o e h

p h 1o

p o o

p o p o p o p o p o

p h 1o

p o

d
W

p o

p o

270 Buy one share of stock. This portfolio requires no initial investment. Maintain this position until time , when the portfolio is worth

A short position in the forward could also be hedged using the stock and money market, but the implementation of this hedge would require a term-structure model.

27.3 Future contracts


Future contracts are designed to remove the risk of default inherent in forward contracts. Through the device of marking to market, the value of the future contract is maintained at zero at all times. Thus, either party can close out his/her position at any time. Let us rst consider the situation with discrete trading dates

. At time Enter a future contract at time , taking the long position, when the future price is , when the future price is , you receive a payment . (If the price has fallen, you make the payment . ) The mechanism for receiving and making these payments is the margin account held by the broker.
b

at times

. The value at time

of this sequence is

Because it costs nothing to enter the future contract at time , this expression must be zero almost surely.

p ~

p o p o Rp o p o go p p o d u Gc f q c d 3 ~ t c c 3 b P ff Vbf b 9 d i d e p p p o p o o o q q

ff VVf

d i

d e

d i

d e

By time

, you have received the sequence of payments

p o

p o

d e

p o

d e

pp o p o o q p o q d e

d i

d e

b 

p o

is

-measurable.

Vp o y wxp o q c c c p 53o

Q 1

4 ig h f

dP

u vc t sr p q

I )

p o

d e

On each

, is constant, so

b V

f f bVf

p h 1o

Deliver the share of stock and receive payment

p h o

W X

p to

p o l
W

p h 1o l
P

W X

q h

p o
c 3

c 3

d e

e d

CHAPTER 27. Bonds, forward contracts and futures


The continuous-time version of this condition is

271

satisfying

Theorem 3.66 The unique process satisfying (a) and (b) is

On the other hand, if (b) holds, then the martingale


U V

satises

this implies

p o 3

p 53o

U e

p o

P q p o

p o 5

i jh i f p 53o

p o 5

p o

p 3o

Proof: We rst show that (b) holds if and only if is also a martingale, so

ui jh h i U Vp o p o 5 f po p o ui j9p i h o

is a martingale. If

i jh p o

p o

Denition 27.1 The -future price of the stock is any

-adapted stochastic process

p o

Note that appearing in the discrete-time version is approximating a stochastic integral.

-measurable, as it should be when

a.s., and

f
(a) (b) is a martingale, then

i jh
c

U V

p o

p o 5

p o 53
P

pto p o e

p o 53

p p53o 5o P d p 5o

p 3o

h d

p o p o

p 5o

d p 3o
U

p o

p o p o d

S T

p o

p o 5

S T

p o

p o p o

p o

p 3o

p o
S T

p 53o

272

and so

is a martingale (its differential has no

With a future contract, entered at time 0, the buyer receives a cash ow (which may at times be negative) between times 0 and . If he still holds the contract at time , then he pays at time for an asset valued at . The cash ow received between times 0 and sums to

Thus, if the future contract holder takes delivery at time , he has paid a total of

27.5 Forward-future spread


p o

Future price:

p h 1o

p h 1o

p o

If

and

are uncorrelated,

p o p o

pp o o

p o

p h 1o

p o

Forward-future spread:

p o

p o

p o

pt o p o

p o

Forward price:
W

pto p o e

p h 1o

p o

for an asset valued at

p o

p o

p ho 1

p h 1o

With a forward contract, entered at time 0, the buyer agrees to pay The only payment is at time .

for an asset valued at

27.4 Cash ow from a future contract


W

p ho 1

p o

p o

p h 1o

pp o

Clearly (a) is satised. By the tower property, is the only martingale satisfying (a).

is a martingale, so (b) is also satised. Indeed, this

f
.

ui

i h

p o po p o e

Now dene

term).

p o t

p ho 13o

p o 3

p o
P

p ho 1

CHAPTER 27. Bonds, forward contracts and futures

273

This is the case that a rise in stock price tends to occur with a fall in the interest rate. The owner of the future tends to receive income when the stock price rises, but invests it at a declining interest rate. If the stock price falls, the owner usually must make payments on the future contract. He withdraws from the money market to do this just as the interest rate rises. In short, the long position and . The buyer of the future is in the future is hurt by positive correlation between compensated by a reduction of the future price below the forward price.

27.6 Backwardation and contango


f p o
C

We have

The expected future spot price of the stock under

is

f g

p o

Because

is nonrandom,

and

are uncorrelated under

p o p o

p h o f p h p o e o q C d e( H p o p o (~ f e p o p o W p p o p o o p o p h o u p q o C p h ( o u p q o p o C ( 7 p o e

p o

Then

is a Brownian motion under

, and

po p to f p H q to q

p o p o p o H p 3o o H p

( p o

(~

Dene

p o

u p o

po

Suppose

p to

f
. Therefore,

p h 1o

i p h H1o

p o

p o

If

and

are positively correlated, then

274

p o

p ho 1

If then

, then

This situation is called normal backwardation (see Hull). If . This is called contango.

p h 1o

p h 1o e

The future price at time

is

po l h p1o

Chapter 28

Term-structure models
C

In a term-structure model, we take the zero-coupon bonds (zeroes) of various maturities to be the primitive assets. We assume these bonds are default-free and pay $1 at maturity. For , let price at time of the zero-coupon bond paying $1 at time . Theorem 0.67 (Fundamental Theorem of Asset Pricing) A term structure model is free of arbitrage if and only if there is a probability measure on (a risk-neutral measure) with the same probability-zero sets as (i.e., equivalent to ), such that for each , the process

Remark 28.1 We shall always have

for some functions

and

. Therefore

275

p o

p o p o p o 5 i p p o o p o p o d

i jh

p o

p o

p o

p o p o q j p o p o d p o i p p o t

i 5

p o p o p o p o 

is a martingale under

i Qi

i h

ho 1e

i h i f ui j9p o i h

Suppose we are given an adapted interest rate process lation factor


Q

ui

i jh

i
H

p5o P

p o p t o

i jp o i h i h 9p o p o
H

Throughout this discussion, , and


W

is a Brownian motion on some probability space is the ltration generated by . . We dene the accumu-

p o

F p h

3o E

276 , the mean rate of return of under , is so is a risk-neutral measure if and only if the interest rate . If the mean rate of return of under is not at each time and for each maturity , we should change to a measure under which the mean rate of return is . If such a measure does not exist, then the model admits an arbitrage by trading in zero-coupon bonds.
H

28.1 Computing arbitrage-free bond prices: rst method


Begin with a stochastic differential equation (SDE)

The solution is the factor. If we want to have -factors, we let be an -dimensional be a function Brownian motion and let be an -dimensional process. We let the interest rate of . In the usual one-factor models, we take to be (e.g., Cox-Ingersoll-Ross, HullWhite).

We showed in Chapter 27 that

for some process . Since and there is no arbitrage.

has mean rate of return

under

is a risk-neutral measure

28.2 Some interest-rate dependent assets


p o f d
d RH

ui

i h

p o

Call option on a zero-coupon bond: Bond matures at time Price at time is

p o

Coupon-paying bond: Payments

at times

. Price at time is

. Option expires at time

uui i

p o

ff VVf

i jh

p o p t o p o p o p p o p o o 

Vp o U

i h p o

Now that we have an interest rate process prices to be

, we dene the zero-coupon bond

p o

p o
n

p o

p o

p o

p Q o

p o p Q o

pt o H p o
n

p o

m 5

p o p Q o

b H

p o

p 3o

f f bVf

sq S ( e po p o p o

p o

k l

p Q o

p o

p o

p Q o

p Q o

CHAPTER 28. Term-structure models

277

28.3 Terminology
Denition 28.1 (Term-structure model) Any mathematical model which determines, at least theoretically, the stochastic processes

for all

28.4 Forward rate agreement


Let be given. Suppose you want to borrow $1 at time with repayment , at an interest rate agreed upon at time . To synthesize a forward-rate (plus interest) at time agreement to do this, at time buy a -maturity zero and short -maturity zeroes. The value of this portfolio at time is

p o

  xo

p o P p t o p 

A q

The forward rate is

  

p s

  xo

p 5

or equivalently,

p f t o

p r

At time , you receive $1 from the -maturity zero. At time effective interest rate on the dollar you receive at time is

, you pay $ given by

3 2 q 3 2 Y

p q

p p5 o

3 2 q 3 2 Y

p o p5

p 5

ui i

p 5 o p p o o q o p

is equivalent to determining

ui i

i jh

i jh

p o o p

p p o p o

Determining

p o

  xo

q d

p t o

or equivalently,

p s o

t o p p

Denition 28.2 (Yield to maturity) For -measurable random-variable satisfying

, the yield to maturity

ui
is the . The (4.1)

i jh q

t o p i uh t(

p t o


p i p 5

l i

ho 1

i h

p o

p o
t

sq p o
t

p o f p o
t

sq

p o p o o t u p u u q u p o u q p o
.

Up o

p 3o

p o p o f q

q p 3o

yp o q p o

Vp o U

sq

S ( e
x 9 w $"v

p o u u e u p o u e p o

d p o

p o

p t o p 5 o

You can agree at time to receive interest rate at each time . If you invest $ at time and receive interest rate at each time between and , this will grow to

p o

p 5 o
t

sq

p t o

u u

p o p 5 o

  xo

p 5 o

u u

  xo

  xo

p o

This is the instantaneous interest rate, agreed upon at time , for money borrowed at time .

Conclusion: at time .

On the other hand,

so

Integrating the above equation, we obtain

278

28.5 Recovering the interest


from the forward rate

p
C

pp  o p

q o

~ }

p o p o o p f p pC o o p o t q o p5 o sq ~ } t o 5 o p q q p 5 o t

p 5 o

p o

~ }

|
C

p o

U |

p1 ko pk1 o p k 5o f } } } e e e p o p o p q o q p o f 2 3 2 C { y y {z | {z p p o q Uz 5 oH q p o S S p o p5 o p o u p o q C p o p o q


t

p o

Now

i jh

53o p
C

p 3o P

p o mq ( t p p u t o t o t uisix 3o h p i i h p sj 3H o p 3H P 1o o p h

p o ho t 1

28.6 Computing arbitrage-free bond prices: Heath-Jarrow-Morton method

CHAPTER 28. Term-structure models

For each

, let the forward rate be given by

Here

and

In other words,

p o
C

p o p o

and Then Let Recall that

~ }

p o

are adapted processes. 279

f
and

p 3o
C

p 3o P

pt o P o p h
t

p o

h 3o p ho i ih p 3oH 1r i jh f ui i ui i f
, let for all

i jh

H i h i f p o(Gt o p pp t o pp t o

p o(G o p pp  o p o p o t o p p C o o p o p q t o p o p p C o o q p o p
under

o q o p o q o p p o


Then

f p 53o P

Not only does (7.1) imply (7.2), (7.2) also implies (7.1). This will be a homework problem.

uui i

i jh

p 5 o

p o

uui i

i jh

p 5 o

i Y eui

i jh

p o o p

Suppose (7.1) does not hold. Then neutral measure. Let

o p o p o H H p3o p3( P o p o q q I ( C P p o o p53( p o C i h o p


C

ui

p H o
Differentiating this w.r.t.

p o 5 H

i.e.,

p o

In order for

28.7 Checking for absence of arbitrage

280

is a risk-neutral measure if and only if

to have mean rate of return

p t o

Differentiation w.r.t.

yields the equivalent condition

Theorem 7.68 (Heath-Jarrow-Morton) For each , be adapted processes, and assume , be a deterministic function, and dene

t i jh 1o i p h i i reh 3o p
t

, we obtain

is not a risk-neutral measure, but there might still be a riskbe an adapted process, and dene

, we must have

and . Let (7.4) (7.3) (7.2) (7.1)

p o(G o p p p t o o

p t o

p o

CHAPTER 28. Term-structure models

281

is a family of forward rate processes for a term-structure model Then without arbitrage if and only if there is an adapted process , satisfying (7.3), or equivalently, satisfying (7.4).

and volatility

. The excess mean rate of return, above the interest rate, is

and when normalized by the volatility, this becomes the market price of risk

The no-arbitrage condition is that this market price of risk at time does not depend on the maturity of the bond. We can then set

and (7.3) is satised. (The remainder of this chapter was taught Mar 21) Suppose the market price of risk does not depend on the maturity , so we can solve (7.3) for . Plugging this into the stochastic differential equation for , we obtain for every maturity :

Because (7.4) is equivalent to (7.3), we may plug (7.4) into the stochastic differential equation for to obtain, for every maturity :

28.8 Implementation of the Heath-Jarrow-Morton model


Choose

p o

p o

p o o u p p o p f p o Q C p (G o u o o p p p

p o p

p o p p p t o o o q

pp o p o t o q p o

ui jh i uf ui jh i i

p o f pp p t o o o q

pp p t o o o q

Remark 28.2 Under

, the zero-coupon bond with maturity

has mean rate of return

i sh p o

p o p p q o p o o o

p o o p

p ( o

i o o

o p

i
H

p o

i gh o p

p o

282 These may be stochastic processes, but are usually taken to be deterministic functions. Dene

Because all pricing of interest rate dependent assets will be done under the risk-neutral measure , under which is a Brownian motion, we have written (8.1) and (8.3) in terms of rather than . Written this way, it is apparent that neither nor will enter subsequent computations. , and the process The only process which matters is (8.4)
C

From (8.3) we see that . Equation (8.4) implies

i h 1o p h

In conclusion, to implement the HJM model, it sufces to have the initial market data and the volatilities

p o

ueui i

i h

p o

p o

This is because vanish.

and so as approaches

(from below), the volatility in

ueujh i i

p o

obtained from

is the volatility at time of the zero coupon bond maturing at time (8.5) must

i mh 1o i p h

and then the zero-coupon bond prices are determined by the initial conditions , gotten from the market, combined with the stochastic differential equation

p o

p uiei h 5 o i p p o i ueui h o i p p H o

p o

ui

p o p o p o q

i jh

p ( o

p o

this determines the interest rate process

p o

p o Q p o t o p

p o

p o

p o

t ui

p o

i h

p o

Then

for

is determined by the equation (8.1)

ueujh i i

1o p h

  xo

p h 1o

uujh t 1o i i p h

Let

be determined by the market; recall from equation (4.1):

f p 53o P

p p
(8.2) (8.3)

o H p o q p53o 53( C p o C o(Gp o Q p

sq I ( p o 5 P o p o
P

p o

p t3o H p o p o C p H o
t t

CHAPTER 28. Term-structure models

283 and satisfy (8.5). We can then dene

and (8.4) will be satised because

Remark 28.3 It is customary in the literature to write rather than and rather than , is the symbol used for the risk-neutral measure and no reference is ever made to the so that market measure. The only parameter which must be estimated from the market is the bond volatility , and volatility is unaffected by the change of measure.
H

i h o p

We then let be a Brownian motion under a probability measure , be given by (8.3), where is given by (8.2) and initial conditions
C

, and we let by (8.1). In (8.1) we use the

p o H p 5 o

ueujh i i

pt o u 1o p h
u

p o p p q o o

  xo

We require that

p o
u u

p o

p h 1o

p o

be differentiable in

p o

284

d b

f f bVf

p Vo

d e

p Q o
d e

u d t


is the row vector

ff Vbf

Denition 29.1 (Gaussian Process) A Gaussian process the property that for every set of times

Gaussian processes

Chapter 29

, , is a stochastic process with , the set of random variables

 p Q p Q o p o o f f bVf b i i i jh i p b Q VVf o f f

Remark 29.1 If

is jointly normally distributed.

is a Gaussian process, then its distribution is determined by its mean function

p oQ p o e

and its covariance function

p o  k k f"ippVo o ~ pVo p o fVVf q f q q d Db b d e k Hp Q ( k Hp Q o o f f bVf b H p Q b p Q o o ff VVf b p oRpV3o o p pV3oQBo p 3o i e p p o f

where

Indeed, the joint density of

is the covariance matrix

f f bVf o p b o p i ff Vbf b b f f b ff p V b f o i Vbf Vbf o p ff ff Vbf p o i Vbf p o ff


b 

, is the row vector

is

, and

D o p

p o

p Vo

i i i

p o ib f f b p bVf o i p o i

The moment generating function is

ff VVf

k k

where

285

p n 1o

t
t

j j

P Q n p n 1o f I )( P Q n p n 1o P t I e p Vo j P n p n 1o wd t j | { Dz y P P n p1o n n p no V po V t o P t pV3o C e p b3o p o b3( t t C

t t

f p 3o

p3o

P 

p 3o
i

p5 P o p o
C

p Q o

h p o

p o 5
S

 S

To prove that a process is Gaussian, one must show that has either a density or a moment generating function of the appropriate form. We shall use the m.g.f., and shall cheat a bit by considering only two times, which we usually call and . We will want to show that

b 

ff Vbf

p Q o

pb3o
C

pp V3o

pb3o pp V3o C p xVo o o p


C i

p o

h p o

29.1 An example: Brownian Motion

286

Brownian motion then

is a Gaussian process with

and

. Indeed, if

i jh

Theorem 1.69 (Integral w.r.t. a Brownian) Let dom function. Then

p o

C C o

f p V3o p C b3o e f V3o p C C j

p o o q C p o

C e p 3o Q e i

Proof: (Sketch.) We have

f
C

p o V3Q

This shows that

t(

Therefore,

is a Gaussian process with

is normal with mean 0 and variance and


Martingale

p o b3Q

be a Brownian motion and

. a nonran(1.1) ,

te

i h i p1o n
:

I )

b3o p

H
o

V3o p

Vo p pb3o p o t Vo p1o n p t

p n 1o
C

p no 1(

h p n o
C

Vo p

t
C

pn1( P o

p Vo

n p n 1o o p n f ot

pn1( o

p o

Now let

CHAPTER 29. Gaussian processes

be given. Just as before,

p o p o f
C

We now compute the m.g.f. for

n 1o p n o p o p Q7V3Q f n 1o P n 1o P p n p o p n h p Q o p o V3Q 5 e P P 5 U U P f P S S ( P P Qn p n 1o p n p n 1o o I )( Q n p1o n n n1o P p p 3o I )( Qn p n 1o f i 4 f t o I7( d 5s e Q pV3o $ 4 tG t I e Qn n1o 4 p t o I)( V3o 4 p i e p o p Q b3Qo p o
(1.1) 1.2

Vo p

te

t t

p n o
C

p Vo t

t t

This shows that

pb3o p o p Vo p o

The solution to this ordinary differential equation with initial time is

pn o

f f bVf

l i sh

to get

Take

Integrate from to to get

conditional expectations and use the martingale property

is jointly normal with

, where

(1.2) 287

288

Remark 29.2 The hard part of the above argument, and the reason we use moment generating functions, is to prove the normality. The computation of means and variances does not require the use of moment generating functions. Indeed,

Therefore,
P n o p n p b3o f o p o p pp o o o p o b3o V3Q q p QBVb3Q p Q7b3Q

If were a stochastic proess, the It isometry says o

However, when is stochastic, is not necessarily a Gaussian process, so its distribution is not determined from its mean and covariance functions. Remark 29.3 When is nonrandom,
C

is also Markov. We proved this before, but note again that the Markov property follows immediately from (1.2). The equation (1.2) says that conditioned on , the distribution of depends only on ; in fact, is normal with mean and variance .

p Q o

n 1o p n f

po n

p 3o

p b3o

p5 P o o p Q

pV3o

i jh

and the same argument used above shows that for

| V3Q p o

V3o Q p

pp o b3Q

{ z b3o p

p o

p QVV3Q o o p o

p n 1o

y p o V

I q

p o V

p b3o

o p o p Q7b3Q

o o p o ppV3oQ q p QBVb3Q i i sjh

by the It isometry. For o

po n

pb3o

p Q o

For xed

h p o p o

p ho 1Q

is a martingale and

, so

p 3o

p5 P o

p Q o

p o V3Q

CHAPTER 29. Gaussian processes


y= v=

289
z
z z

s y t (a)
y=

s v (b)

s v y t (c)

Proof: (Partial) Computation of : Let be given. It is shown in a homework problem that is a jointly normal pair of random variables. Here we observe that

and we verify that (1.3) holds.

n x 3x p o p o f

p o 53Q

p3x P o

i h

p n 1o

p o p o e

p 3o

zi

p 3o

p o p o p V3so

Then

is a Gaussian process with mean function

and covariance function

p x o

p5op53o P

p ( o

h p o

p o

5o p

p53( P o

p o

p o

Theorem 1.70 Let Dene

be a Brownian motion, and let

and

n (

Figure 29.1: Range of values of

for the integrals in the proof of Theorem 1.70.

be nonrandom functions.

(1.3)

n p3 3x o p o n p o 3x p o

p oxp3 o

p x3 o p o

p n o

p oxp3 o

p o

p n p o 1o G3x

p3 1o Gp o p n o
P

p o
P 

p n p o o G3x

px 1o Gp x o p n o
P

p n 1o n n

p x o

p o 3x

p1o n

px p o o

po n

p 3x o p o

P P o p o n o p n p 3x P P P o p o n 1o p n x p 3x p 3oQ o p p 3x P P o p o p Q7p3oQ7p x3x P P o o p o e p oQ7p o P p3oQ73x P f p o e o p o p 3o p V3 Q e

Remark 29.4 Unlike the process

We have

290

, the process

p o 3Q
P

p s o

p 53o

p o 3(
P

p Q o

(See Fig. 29.1(c))

(See Fig. 29.1(b))

(See Fig. 29.1(a))

is

p n 1o

P P P

p n 1o p n 1o
P

p n 1o
P

p V3o

b3o p

neither Markov nor a martingale. For

where we have used the fact that is a martingale. The conditional expectation , nor is it a function of alone. not equal to

p o

p o V3

p o V3

CHAPTER 29. Gaussian processes

291

is

p3x V3Q b3s o p o p o

pV3oQ7p3ox p o b3s

p V3o

p o 3Q

p o 3x p o b3s p o p o 53Q73x l i jh
P

p53oQ7p3x o

p Vo

p s o

292

P V

p3o o p 5o

~ 1

~ p 3o e
i

P p h o

~ p o e

3o p
C

p5o o

53o p

p53o 

p 3o 

P p h 1o

p o

Integrating, we get

e e

p p o p o u p o 5 o f p o C p o p o f

7p o


Then

p 3o

p o p o p o
where , and

p o
C

Consider

p o p p o o p o u p o p o q p H1o
are nonrandom functions of .

so

p o p o 53H

P p h 1o

~ p o e

The process

and covariance function

From Theorem 1.69 in Chapter 29, we see that

We can solve the stochastic differential equation. Set

Hull and White model

Chapter 30

is also Markov. is a Gaussian process with mean function (0.2) (0.1)

293

p o

n f

p1o n o

p o 3H

~ P

p ho 1r p h t 1o
where

t 1j 1o 1o p ho p h p h q q d( P Q n ~ p n 1o  ~ P p5o  p h 1o ~ P P q ~  q I ( P p o P pd qo p o pd qo (

x&

p o

p h 1o

n f

p po 3o 7 o p Q1 o

P P P p3oH p h p o 1o ~ o g x p o P P p h 1o w

p o f

The price at time 0 of a zero-coupon bond paying $1 at time

p 5o 

P p h o

P ~ P p o 9 ~ p o e

p o

3o p
C

P p5o o p Q o p o P

and its variance is

According to Theorem 1.70 in Chapter 29,

Then

We want to study

294

. To do this, we dene

is normal. Its mean is

is (0.3)

p1o n 

po

p o

&

Because is a Markov process, this should be random only through a dependence on

p o

p oj o p o p p o q q t( UVp o p53o p o f sq ST e h pt 1oj pt h1o p1o h h p h 1o f q q ~ P p h o p1o n  q


of the zero-coupon bond:

p no 1H

n f

P ~ p no 1H p n w 3o  ~ P ~ p3oH o p53oH ~ P P

CHAPTER 30. Hull and White model

Figure 30.1: Range of values of

Consider the price at time

Therefore,

Note that (see Fig 30.1)

30.1 Fiddling with the formulas

for the integral.

. In fact, 295

p ho r

u = t

p o
C U V

p o p o p o p o bVf f f

p o

p o

o p f p o p o q p p o p o p o p o p o p o t o p q

p h 1o p o p o p 3o uq p 5o

p o p n 1o q
P

p 5o

p n 1o

The reason for these changes is the following. We are now taking the initial time to be rather than should be replaced by Recall that zero, so it is plausible that

n f Vbf ff

VVf P ff

f n
x

p1o n o q

p n 1o

p o p r o
where

Similarly, see that the

and this should be replaced by

296

should be replaced by terms cancel. In

. Making these replacements in , however, the term does not cancel.

p ho 1j

p o p 3o

Let

30.2 Dynamics of the bond price

to obtain the bond price, its differential must be of the form

p 3o

S (1T e

p o

Because we have used the risk-neutral pricing formula

q p o C p o q q p o

p o p o p o p o op o p o

q q p o q p o p o

t or o p o p p p t o q q t( p o p t o
and

we have denote the partial derivatives with respect to . From the formula , we

p h 1o

p o p

h o j

po n o q

p o

p o p u p o p o

p o p H o p h 1o p o p o p o q q t ~ p o p n1o p r o o ~ 3o P p p o p o 1o p o q

p o
C

p t o

Therefore, we must have

We leave the verication of this equation to the homework. After this verication, we have the formula

p o p o p o p o p p q o p o o 
.

p o

p o

p o p o p o Bx o p t o p o p o o p p o p q q q q

Suppose we obtain determine the functions

30.3 Calibration of the Hull & White model

In particular, the volatility of the bond price is

CHAPTER 30. Hull and White model

Recall:

for all , , and

from market data (with some interpolation). Can we for all ? Not quite. Here is what we can do.

Step 1. From 4 and 5 we solve for

We take the following input data for the calibration:

1.

ueuh 1o i i p h

2.

p h o

3.

uiuijh p o o h p h ui ijh p o p h 1o
; ;

5. 4.

(usually assumed to be constant); ; 297

, i.e., the volatility at time zero of bonds of all maturities.

k s!

h h d p k 1o p5khh h do k s! h 51o p k
}

n f

po n o q q p h 1o

1r p ho

P po p 3o f u o u p h
u

q e

p o p o

h e u
u

  

p h t 1o

u
We can then compute

Step 2. From the formula

p h 1o

P ~ p no 1H p ho 1j o ~ p ho t 1j h e 1o p h

We can use this formula to determine

p o q p o p o p f uieuh po p ho 1r u i pto o p p o f u q P n p1o 1j u n p to p ho o q u P n 1j u ~ p1o n p o 1 p ho  q u P n x U ~ p no p no p ho 1j u o q 1HP  u ~ 1S P n " ~ ~ pno U p no p ho 1j u o q ~ 1H  ~ u 1S uujh to i i p

From assumption 4 and step 1, we know all the coefcients in this equation. From assumption 3, . We can solve the equation numerically to determine the function we have the initial condition .

This gives us an ordinary differential equation for , i.e.,

We now have

p to

we can solve for

298

Remark 30.1 The derivation of the ordinary differential equation for requires three differentiations. Differentiation is an unstable procedure, i.e., functions which are close can have very different derivatives. Consider, for example,

p o

p ho 1H

i jh p H o

o
u u u u

for all

for all

. Recall that

as follows:

known function of

u u

P p53o p p o o  4 4 ~ f e x& P 4 p53oH7 ~ p h 4 ~ 1o p o 4  e p o

n f n

~


P 4

p1o n 


P 4

p o 3H

P 4

~ 1o p h p 3o

p k  o f p
t

p o  r p o q q D r p o p q
.

Then

Assumption 5 for the calibration was that we know the volatility at time zero of bonds of all maturities. These volatilities can be implied by the prices of options on bonds. We consider now how the model prices options.

Consider a European call option on a zero-coupon bond with strike price The bond matures at time . The price of the option at time 0 is

h d p5khh o h d k s!
h d

p ko 5
}

p ko 511
t

p k 51o
}



h h d d

i 51o p k
}

p k 5o
t

for many values of .

we have

but because

CHAPTER 30. Hull and White model

30.4 Option on a bond

and expiration time

We observed at the beginning of this Chapter (equation (0.3)) that

p 53o

p53o 4 P p o ~ q ( o p o o z o q p p o o x q ut

We also observed (equation (0.1)) that

5~

5~


4 

j ~

where

is the joint density of

is normal with

U z

p 53o

P 4

S x&

U z

p 53o

P 4

is normal with 299 .

p k 1o

Because of the Markov property, this is random only through a dependence on . To compute this option price, we need the joint distribution of conditioned on . This

p o

p o

p o

p 3o

p o

j p o q

p  o

k
S

p i o (i x q d q d q d d  r D o p o p

P p 3o i 4 P pp o p o H5o o pp p 53o o 4 q q UVpp o p o o 3o pp p 5o o P i f q 4 S T e

The option on the bond has price at time zero of

The price of the option at time

In fact,

7p o

p 53o

where

300

is dened in Equation 0.2.

is jointly normal, and the covariance is

is

(4.2) (4.1)

p o o  4 o q p o p p D q

~ e o o 
o

5~

5~

d(

p 3o

U e

p o

p 5o

sq S e

p o

y U V

f p o p 

n q

p n 1o p o o

p o

p 5o 

p o n

p 3o

P4 ~ ~ 4  o ip o p 53o p  o q 4 S ~ e p 3o o 4 4 ~ e p o pp  o p o o p o q e 4 ~ 4 ~ p o p o p o p  o e ~ p1o n 4  4 ~ vp o p5o p o q 4 ~ e ~ ~ p o 4


U V 

p o

p 53o

pair of random variables has a jointly normal conditional distribution, and

Advantages of the Hull & White model:

The variances and covariances are not random. The means are random through a dependence on .

CHAPTER 30. Hull and White model

Short-comings of the Hull & White model:

2. Interest rate is normally distributed, and hence can take negative values. Consequently, the bond price

r p o q

1. Leads to closed-form pricing formulas.

2. Allows calibration to t initial yield curve exactly.

1. One-factor, so only allows parallel shifts of the yield curve, i.e.,

p o p o p o q t(

p o p o
i

p o e

can exceed 1.

so bond prices of all maturities are perfectly correlated.

p o

301

302

Chapter 31

Cox-Ingersoll-Ross model
In the Hull & White model, is a Gaussian process. Since, for each , is normally distributed, . The Cox-Ingersoll-Ross model is the simplest one which there is a positive probability that avoids negative interest rates.

is called the Orstein-Uhlenbeck process. It always has a drift toward the origin. The solution to this stochastic differential equation is

This solution is a Gaussian process with mean function

Dene If

, we have
I 'H

There are innitely many values of 303

for which

Qh p o h Q hQHp o p o d p o H p o Vfbf p o p o p o f f P p 3o i ~ d

and for each ,

and covariance function

, but (see Fig. 31.1)

53o p

p o

c C

c C

p o

P o p h

p h 1o

p o e

p o q

~ p o e

and let

be the solution to the stochastic differential equation

p h 1o p h 1o 1o p h ff Vbf

We begin with a -dimensional Brownian motion constants. For , let be given so that

. Let

and

p o

ff Vbf

h l QHp o p h Ho

p o

VVf d ff

be

u t

p 53o

p 53o P p 3o

p o
C

u t

p o

p o p o

p o

u t

p o


u t t

p o q ~ p o u t Q q q u t
u t t

p o
It s formula implies o

88

h f

h k 8 $8 t t k VVf k k ff h Q p o
. Then

ff VVf

h p o

H
If

Let

304

, (see Fig. 31.1)

There is at least one value of

r(t) = X2 (t) 1

Figure 31.1:

x2

can be zero.

for which

if if

p k

k 1o k

Dene
( X (t), X (t) ) 2 1 x 1 t

CHAPTER 31. Cox-Ingersoll-Ross model


Then
C

305

is a martingale,
C

so

is a Brownian motion. We have

The Cox-Ingersoll-Ross (CIR) process is given by

If

happens to be an integer, then we have the representation

This is not a good parameter choice.

p ~

o p o q d

For

is normal with mean zero and variance

p h o

p h 1o

p h 1o

h Q

~ p h o

f f bVf

p o

For example, here is the distribution of

for xed

. Let

be given. Take

With the CIR process, one can derive formulas under the assumption that integer, and they are still correct even when is not an integer.

h Q

There is at least one value of

for which

h p o

p o d q bf d ff p h 1o h o p h

If

(i.e.,

), then

h u

There are innitely many values of

for which

f d

h p o

but we do not require

to be an integer. If

(i.e.,

p o

h Q

u t

p o

We dene

), then

p o

p o

p o

p o

p Q p o

p o

is a positive

1o p o

u t u t

p o

306

and variance

. Then
p | o { z y

Normal squared and independent of the other term

Chi-square with freedom

degrees of

31.2 Kolmogorov forward equation


Consider a Markov process governed by the stochastic differential equation

p o

p p oo p o p Qo

p o

We computed the mean and variance of

in Section 15.7.

p o

We make the change of variable

. The limiting density for

p o

{ d

p o

p o

p o

t u

As with

, we have , and so the limiting distribution of is times a chi-square degrees of freedom. The chi-square density with degrees of freedom is

is

p o i p o

u t

p o

p o

p o

h up o

As

. We have

x 9 w v

31.1 Equilibrium distribution of


p o

Thus

has a non-central chi-square distribution.

p h 1o
(0.1)

p o ~

~ { p Do u z y i p o i p o ~ Dt ~u

p o p o i

p o

is normal with mean

P ppV3oQo p p o # p p o p p o pp ho o p o o 1Qx p x b3Qo bgQo m V3Qo b3o p pp o pp o Vo b3Qo P C pp o V3Qo V3Q#` VQo m V3Q( P 1Qx p Qx pp o o pp o pp o o pp ho o p o o

p o
C

p p oQo p o p o

p p oQo p o p o p p Qo # p o o m p Qo
which vanishes near

p o o p Qx

p  o 3 P p o p o o p x k p Q o h k

p o h p k 1o

We start at at time 0. Then is random with density variable). Since 0 and will not change during the following, we omit them and write than . We have

Because we are going to apply the following analysis to the case for all .

CHAPTER 31. Cox-Ingersoll-Ross model

Figure 31.2: The function

, we assume that

(in the rather

p o

p Q o

p o x

Let be a smooth function of Fig. 31.2). It s formula implies o

p o 3

for any function .

p k 1o h

p ho 

h p Q o

so The Kolmogorov forward equation (KFE) is a partial differential equation in the forward variables and . We derive it below. and for all large values of (see 307

f
at that and nearby

Uep o

q p

im

u u

u pp p o p o po u q  o 3o m o

u pp p 7 o p3o p u q  o 3o m o

u p o p  o p3o x P u

| p p o  o p3o u x P P u f
u u

p o p3o

pp  o 3o o p
m

p  o p3o 3o # P p

If there were a place where (KFE) did not hold, then we could take points, but take to be zero elsewhere, and we would obtain
u u

h j3x p o

pp  o 3o o p
m

u u

{ Dzu p )  o p3o

u | P p3o P sq u u

|u P p3o P q
u

{ Dzu y p o 3o p p 3o P y { Dzu p  o 3o 3x p p o


p o p3o p3o P
m

p p  3o 3o 3o # P P p p 3o p3o p3o P P 1Qx pp ho o


m

This last equation holds for every function


u u S

p  o

p3ox P q

p  o 3x P p o
Therefore,

u u

p o 3x q

p  o 3o 3o # P p p p  o 3o m 3o P p p

p  o 3x P p o

p  o 3x P p o
or equivalently,

or equivalently,

Integration by parts yields

Differentiate with respect to to get

308

of the form in Figure 31.2. It implies that

(KFE)

p  o

p o 3 P

u u

u u

p o

p o p 1o q q p o p q q q

We want to verify the equilibrium Kolmogorov forward equation for the CIR process:

~


f d h

p 3o

p 3o

u u

u p p pp p o 3o 3o o m o u q

p o

p 3o

p o

If the process

p Q o

We compute

p o

Letting

t Q

where

31.3 Cox-Ingersoll-Ross equilibrium density

When an equilibrium density exists, it is the unique solution to this equation satisfying

In order for this limit to exist, we must have

CHAPTER 31. Cox-Ingersoll-Ross model

We computed this to be

in (KFE), we obtain the equilibrium Kolmogorov forward equation

has an equilibrium density, it will be

(EKFE) 309

p p o o

p o q p o p q 1o

q p # o

q o ! p o

p 1o q q d q q p o q q p o ! q p o f

q d q

p o

p o

the tower property implies that this is a martingale. The Markov property implies that is random only through a dependence on . Thus, there is a function of the three dummy variables such that the process is the function evaluated at , i.e.,

p o

p o

p p o o p t o f p o p o p o p 53o
P

Up o

U V

p o

p o

p 53o

sq T e S

p o

p o
C

p o

p Q p o

1o p o

p q

p q p

p o

p q 1o q q q p 1o q q p 1o ! p q q q 1o p q q q q 1o d p q q q t q p o p o p q q

p # o o p f u p o p Q p o o u p o o u p u u p o p p p o p o 17o q

31.4 Bond prices in the CIR model

1o d q q 1o p o

1o d 1o 5p q 1o p q

1o p o

as expected.

The LHS of (EKFE) becomes

Now

310

The interest rate process

p o

u u

where

p h 1o

Because

p5o P q
I

is given. The bond price process is is given by

p t o

p 3o j p o p `h p t o o q d q p t o

p 3o p o q q d q h p o p o q q q q q 7 q p o q q 2

~ 2 (~ p o } e d

Surprisingly, this equation has a closed form solution. Using the Hull & White model as a guide, we look for a solution of the form

ul ijh h f h t o P o p p p 1o p p t o t o q q
Setting the term to zero, we obtain the partial differential equation

7 f p 1o q q
equals

p o o p p o f p o p p o o pt p o o t p

p o p p o Q q p o o R Q

p o o 7 p53o P

p3o P q

q
d

Because pute

p p o o

The terminal condition is

f
(4.1)

p t o

bVf f f

sq
I

( (

I )

p 3o P

where

p j h o

p o

and then set

We rst solve the ordinary differential equation and the partial differential equation becomes The expression in

CHAPTER 31. Cox-Ingersoll-Ross model

. Then we have is a martingale, its differential has no term. We com311

p p h d h o o p p h 1o o
P

p 5o

p o

 p

h h p p o or q p o q t( p t o p j o q p o p o p o p p o 1o p o u q
C

p o o p

U 3

p o

p 53o

T S
Thus in the CIR model, we have

f ~ p p q o o

o j

h p t r o

9 ~  q  p p q o o p q (~ $   xo p p o o p o o p p t q q p p o o   q p o f p o q

where

where

ul

i jh

j p o

p o

t(

do not depend on , the function . Similarly, and , and we have

p j o p o

and

It is tedious but straightforward to check that the solutions are given by

so

312

and

and

are given by the formulas above. Because the coefcients in

depends on and are functions of

only through their difference . We write instead of

p o

p o q

p o

where

f 9 p o p j o q    xo o p o p o  p  o 

We have

p p h 1o o

Now

for each , almost surely, so

is strictly decreasing in . Moreover,

h p Qw5o

CHAPTER 31. Cox-Ingersoll-Ross model

313

But also, so

and

is strictly inreasing in .

31.5 Option on a bond


The value at time of an option on a bond in the CIR model is

where is the expiration time of the option, is the maturity time of the bond, and . As usual, is a martingale, and this leads to the partial differential equation (where .) The terminal condition is

Other European derivative securities on the bond are priced using the same partial differential equation with the terminal condition appropriate for the particular security.

31.6 Deterministic time change of CIR model


p o
Process time scale: In this time scale, the interest rate equation

is given by the constant coefcient CIR

Real time scale: In this time scale, the interest rate

is given by a time-dependent CIR equation

i ui

i h

Vp o U

p h p o o p !n o f q p !n o n 7 ul ijh h h n p n 1o n n f q q p p o d 53o P o n f p

f h

p p o p o p p o zp o p H o o p o o q f C p o p o p o p p o Q p o q 1o

por po p1o h p p h 1o o q q t( p p h 1o o f q


Q

p  o

t o p p h pr o 1o h por p h p h h 1o 1o 

p3o P sq
R

ptor o p

( p 5o
4

p h o

S ( e

p p o dn o

314

A period of high interest rate volatility

Figure 31.3: Time change function.

We need to determine the relationship between and . We have

and this will be

if we set

p p o o o p

p o

p o

p o

p p o o

p o

p t h 1o o p h

p h 1o o p h

p o

With

, make the change of variable

p o P p h 1o o p h sq f P p o p t h 1o o p h

p o

where

, and

and

are as dened previously.

3 2

denote the price at real time of a bond with maturity Let is . We want to set things up so

p j o p o p t o ~ 2 1 3 (~ p o p o } e

p o

There is a strictly increasing time change function Fig. 31.3).

"

which relates the two time scales (See when the interest rate at time

. . . . . . . . . . . . . . . .

p o

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

"

p o

Process time

Real time

in the rst integral to get

p7! h ph1o o l Hp h 1o o p h q q   xo   xo h 1o p h l H h 1o o t l p p h i h q    t p ho 1j p h p h o o f p ho h r h o  o p h p h

p p o 1oj  o 1o h pp h

p o pp o h1o oj  o 1o o p pp p h

p1o h p h 1o

p h o o p h

  

q
or equivalently,

fgp h p h1o h p h 1o o p h 1or p1o q q d p h 1o o p h p h 1o h q pp H o p o j o p p o p p p o p o o

o j o

pp H o p o j o

fgp j o p q p o p q p o p o
i

p o

o p o q d e( p o o q ( p o p o p p o o p o ~

do not depend on and are time dependent.

where

31.7 Calibration

CHAPTER 31. Cox-Ingersoll-Ross model

Suppose we know

and

only through

for all

Take and the functions equation for

The right-hand side of this equation is increasing in the having limit at , i.e.,

so the equilibrium distribution of seems reasonable. These values determine . Take (we justify this in the next section). For each , solve the :

, since, in the real time scale, the model coefcients

. We calibrate by writing the equation

variable, starting at 0 at time

p o

p h 1o

p h o o p h

  

p h o

p o

Since is

(*) has a unique solution for each . For

so

p  o

l p o

. If

. Thus

, then

is a strictly increasing time-change-function with the right properties. , this solution 315 and (*)

f p o p o p o j q  $   Y xo p p t o o p o  o p 

r o p o p h

where , determined by market data, is strictly increasing in , starts at 1 when and goes to zero as . Therefore, is as shown in Fig. 31.4.

p h 1o o p h

  xo

t Q

 o j H o o pp o pp

p h 1o

p h 1o d

p h 1o o p h

p h 1o

h 1o o p p h

  

In the real time scale associated with the calibration of CIR by time change, we write the bond price as

h o o p p h

p h 1o

t 1o p h
u u

u u

p h 1o p h 1o

  xo

  xo u u

z P

f x ~ o j p ~ o

p 5o

q p h 1o f

Hxo  
P

  xo

p 53o

q
P

( e 1o p h
u

p h 1o
  

31.8 Tracking down

x w $G)

Consider the function

p h 1o

p h 1o o p h q h

  xo

The calibration of CIR by time change requires that we nd a strictly increasing function such that

Here

thereby indicating explicitly the initial interest rate. The above says that

Justication:

316

Result for general term structure models:

and

are given by

p o r

p o

in the time change of the CIR model

(cal) with

p h 1o

CHAPTER 31. Cox-Ingersoll-Ross model

317

Strictly increasing

Figure 31.4: Bond price in CIR model

Figure 31.5: Calibration The function is zero at , is strictly increasing in , and goes to as . This is because the interest rate is positive in the CIR model (see last paragraph of Section 31.4). To solve (cal), let us rst consider the related equation

(cal)

t p u o

t Q

p o

l p o

If , then . If dened a time-change function than (cal).

, then . As , . We have thus which has all the right properties, except it satises (cal) rather

p o r o p

p h 1o

p h o o p h

  xo

p o

p o

Fix

and dene

to be the unique

for which (see Fig. 31.5)

pp o oj p o o p

.... .... ..... ..... ..... ..... ..... ..... .... .... .... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

poj to p

p h 1o

p h 1o

p h 1o o p h

p h 1o o p h

p o

ptor o p

  xo

  xo

q p h 1o

Goes to

p h 1o o p h

t u

  xo

h f q p ho 1 q

7p o

p o $

p p h1o q h o 5 7p o p $ p o 

p o

U q d S p h o d h o q p o $ d p o S

p o q

p o $


p o

p o

p o

p ho p p ho 11 15h q h o f d  d p o p o q  p o p o

$ 

f d

p h 1o

p h 1o d p h 1o 1o p h

p h p h 1o o p h o f h 1o p h p h p ho 1o1 11 1  p ho p ho f ph1o ph1o o p h p p p 1o 1h o o P p h 1o o p h


so also satises (cal). From (cal) we compute , so we have
u

u u

  xo

We conclude by showing that

p h 1o

p h 1o

Computation of

p h 1o

Computation of

p h 1o

Note that is the initial interest rate, observed in the market, and is striclty positive. Dividing by , we obtain

We show in a moment that

318

: : ,

p o

p h 1o

p h 1o

 (i

p o

p o
C

p o i q d p o i p o f C o C C p o p o p
C C

p o p

q d

p o i o p o p C p o p o p
C

o o

k k

u p u p

m 5

p o m5 p o

p p o o k p p o  o k

u
P

A two-factor model (Dufe & Kan)

Chapter 32

Let us dene:

Interest rate at time

p o

p o

Let , differential equations

be given, and let

Yield at time on a bond maturing at time

and

p Y o

p o

h 1 o p ho h p h

where

and

o o

and

Then

is a Brownian motion with

are independent Brownian motions. To simplify notation, we dene

319

be given by the coupled stochastic

(SDE2) (SDE1)

p f

i o

hQ o 1o p h p h
will stay strictly positive provided that:

p o i p  C p m q
m


k k


k k

p  o
C

p D m

p o C p  m o m D p (i q C D
k

p o p p 7D k d C o
k

is a Brownian motion. We shall choose the parameters so that:

(i (i d p q

p i o o
C

(i o k i p m

7p s ok C D k

o  k   k

o o

32.1 Non-negativity of

where

320

Assumption 1: For some ,

and (SDE1) and (SDE2) make sense. These can be rewritten as

From our discussion of the CIR process, we recall that

p m q m (i o 7

p o

Then

Assumption 2:

p ho 1

o o #

and

Assumption 3:

Under Assumptions 1,2, and 3,

almost surely,

(SDE2) (SDE1)

p o f p o
C C

p o p o

k k

p u D u p

m 5

p  o m5 o p  l t i h

p p o o k p p o  o k QHp s h o

o o

p o o p k o k r q p  5k q p o k q t(
(PDE)

h f p x

p 

m s

4 px k p k o x k k o ( i 4 5 k p k k k o '  k o p ' 5 k  o q k q m m k k k k p 5k 1o k 4  u  u C C 4 4 4 i 4 P D p o p 5   o q Q 3o sq m I ( k k k k 4 4 4 4 Q 3o P p sq q q I ( p o p o o Q p
is

p53o P sq

p o p o o

p53o P sq
I

f p 5k 1o k

U V

p o

p53o p p oV p o o q q 1S e k k q

Since the pair of processes is Markov, this is random only through a dependence on . Since the coefcients in (SDE1) and (SDE2) do not depend on time, the bond price . Thus, there is a function depends on and only through their difference of the dummy variables and , so that

p o

p Y p o o

U V

p o
is

p 53o

sq

S (1 e

p o

The value at time

valid for all


I

We seek a solution of the form

The partial differential equation for

is a martingale. We compute its stochastic differential and set the

The usual tower property argument shows that

32.2 Zero-coupon bond prices

CHAPTER 32. A two-factor model (Dufe & Kan)

and all

of a zero-coupon bond paying $1 at time

satisfying term equal to zero. 321

(*)

h k

k k

5k 1o k o k k p q k o k o k k q

p o

m s

p o p oV p o  ( p o i p Y m p o m p o p o q q p o o p Y p o ( p o i p o  p o p o 5p k q q k k p o p oY p o ( p o i p  o p o  p o d q q k k p p k o 5 p p k oY p o px k k o (i p o 5 p o  s  p m k k p k p o k k
k

o k o q p p o
k

k 5k 1o k o q k 5p k h k 1o
(PDE) becomes

k 1o k 4 k 1o k k 1o 4 4 k k 1o k 4 k 1o k k 1o k o p o

p 5k 1o p o k p f p k 1o p V p o p k o p 5k 1o p o p k p k o p Y p k o p k o p o q p k p 5k k1o p o p o k p o k q p q q q p j p h Q o h por 1o h p h p h 1o f k k p h k 9 1o k !


. This implies the initial conditions satisfying (*)

We must have

We get three equations:

We want to nd

because

322

corresponds to

for

. We have

(IC)

(3) (2) (1)

p o q p p o q p o

(h o o b p o i q p o q p  5 p o m m (h p o o (i q p o q p Y  p o k (hl k d
k

p ho 1r p o p1o h p o p h 1o p o

p o o q p b p o (i q p o q p o
k

p o 

CHAPTER 32. A two-factor model (Dufe & Kan)

323

We rst solve (1) and (2) simultaneously numerically, and then integrate (3) to obtain the function .

32.3 Calibration
p p oY p  p o p o o p o o P oj P p b p o q P q P q
P

and the yield is

This equation must hold for every value of

and

, which implies that

We must choose the parameters

so that these three equations are satised.

Y i w b m 

But we have set up the model so that Thus

is the yield at time of a bond maturing at time

h h p o o r P p P  f p  o p o p j o o p P Y p o p P o p o f P p o P p j o o p p P Y p o p P o p o f P d

o p V p o o

p Po

p  o

  xo

P q d

h Q

Let

be given. The value at time of a bond maturing at time

p j o

is

324

Chapter 33

Change of num raire e


Consider a Brownian motion driven market model with time horizon . For now, we will have one asset, which we call a stock even though in applications it will usually be an interest rate dependent claim. The price of the stock is modeled by

where the interest rate process and the volatility process are adapted to some ltration . is a Brownian motion relative to this ltration, but may be larger than the ltration generated by . This is not a geometric Brownian motion model. We are particularly interested in the case that the interest rate is stochastic, given by a term structure model we have not yet specied. We shall work only under the risk-neutral measure, which is reected by the fact that the mean rate . of return for the stock is We dene the accumulation factor

The zero-coupon bond prices are given by

325

U V

p o

p 53o

p o

p o

sq

po p o e

1 e S

p o p o p o

p o

p o p o

so that the discounted stock price

is a martingale. Indeed,

i p o h

p o

p o

p o p o u p o
Q R

p3o P

I )

p o (

p o p o p o

p o
C

(0.1)

i h p o

326 so

of the stock is the price set at time for delivery of one share of stock The -forward price at time with payment at time . The value of the forward contract at time is zero, so

Denition 33.1 (Num raire) Any asset in the model whose price is always strictly positive can be e taken as the num raire. We then denominate all other assets in units of this num raire. e e
Example 33.1 (Money market as num raire) The money market could be the num raire. At time , the e e stock is worth units of money market and the -maturity bond is worth units of money market.
g 3 g

We will say that a probability measure is risk-neutral for the num raire if every asset price, e divided by , is a martingale under . The original probability measure is risk-neutral for the num raire (Example 33.1). e Theorem 0.71 Let be a num raire, i.e., the price process for some asset whose price is always e strictly positive. Then dened by
 H

is risk-neutral for

Example 33.2 (Bond as num raire) The -maturity bond could be the num raire. At time e e is worth units of -maturity bond and the -maturity bond is worth 1 unit.
H

p o

p o o p u

p ho 1u

p o p o

H

Therefore,

p o

p t o o p p o W q p o pt o p o p o W q p o p o po o p o p p o p Wq W o pto p o Rt o pp pto o h p o e

is also a martingale (tower property).

W a

p o

p o

p o

e p o
H

p o

H

p o p o

e 

, the stock

  

 

G3

G3


Fix and let
h vh3Xr

H

p o

p o p o ptu o o p p o p o s tu

p o  p s o p o  p o p o  p o

p o

p o  p o q po p tou p o f po 1ou p h p u  o p o pu o p h p o o  ui jh i i

p o

Let be an asset price. Under , is a martingale. We must show that under , is a martingale. For this, we need to recall how to combine conditional expectations with change of measure (Lemma 1.54). If and is -measurable, then

d p h 1o o  p h p ho 1u d p h o  p o p u o f d E p o p h o  p E 3o o p u d
# 

H

p t3o
H

p o

H

p u o p o

p ho 1u

Proof: Because

and we see that

Note:

CHAPTER 33. Change of num raire e

and

are equivalent, i.e., have the same probability zero sets, and

is the price process for some asset,

is a probability measure.

H

33.1 Bond price as num raire e


#

p o

which is the martingale property for

Therefore,

be the num raire. The risk-neutral measure for this num raire is e e


under

. is a martingale under . Therefore, 327

po  p o ts

328

Because this bond is not dened after time , we change the measure only up to time using and only for .

, i.e.,

This is a martingale under

, and so has a differential of the form


5

33.2 Stock price as num raire e


Let be the num raire. In terms of this num raire, the stock price is identically 1. The riske e neutral measure under this num raire is e
e

Denominated in shares of stock, the value of the -maturity bond is

This is a martingale under

, and so has a differential of the form

G$

8 Y6

@ ' S

G$

8 6 $a

Theorem 2.72 The volatility words, (2.1) can be rewritten as

Q `

G sG$T E 1 1 X 2W

where generality that


e

is a Brownian motion under

. We may assume without loss of

in (2.1) is equal to the volatility

 

 (

HI @ S V e

G$

T U

x3

S e

Q R

H I

H P

We write

rather than

from now on.

G 8 3h$6 @ E 1 1 FD( B

i.e., a differential without a term. The process . We may assume without loss of generality that

1 1 35A

 1 1 45352

C @ '

8 6 97G$

 )

is called the stock is

-forward measure. Denominated in units of

 (

#  # "  % '& $  ! #  % "  H ) @ B

-maturity bond, the value of the

(1.1)

is a Brownian motion under

(2.1)

in (1.1). In other

(2.1)

x x x x

0 0 0 0

h3

p q A o VElyk3 B G ri 5 n H FI ) E lyk3) B G WmEly3 B i k H 0 5 0 I Ely k k ) B 5 G WimEly ) B I Hd 0 d  x h f e# xgh%  H h f e# xgd$  H


G3

Wi

d gi

h f e# jgh%  H

Wi

d 9' 0

h f e# xgh%  H

i $WI

 5

d 8 6

Under is a Brownian motion. Under this measure, has volatility and mean rate of return . The change of measure from to makes a martingale, i.e., it changes the mean return to zero, but the change of measure does not affect the volatility. Therefore, in (2.1) must be and must be

G$

8 Y6

# 

H P 8 # 

G$

r )

G$

 @' S i Q  hY6ye 8 v i 8 6 h$a )  @ V w 2i G$ r8 6y5i v 8 hY6xi ) e t e r @ V v5e G$ h$6 8 i p r8 6 ) r   ) e i he ff b  ) 5e v Gi f u b ) ) ) ) ) b S


Ge

) t s c f b r gc ahf g39c phgb i cf e e q

33.3 Merton option pricing formula


v 5e @ i e H H @ @
G$

e X

Proof: Let

CHAPTER 33. Change of num raire e

The price at time zero of a European call is

8 6

r8 %6

i V

Q  )

c d1b

, so

. Then 329

 i r 5h3  x3
is not constant, we still have the explicit

Px

r 3i r8 6 pxhrv  i r 3i r8 6 hrv  v

r  x3 r8 6 i 

G3 x3

r r8 6  i r 3 r8 6 

r 5a 8 G i{ ~ } $6 o 0  @ i  x3 n 0 ~ } i{ 8Y6 x3 v k  @  G n 0 ~ i}{ r x3 @ k r8 6  iP 8 $6 E k Fly3 B G3 5 n ) v us @ w gG3 0 

r i r8 6  I

r r8 6  v

G3 x3

 5a r h3 ~i}{ 8$6 o 0 8 6 v   @ x3 n zH  H ~ } S o p h3 i{|o 8 r @ 0 r 6  iI 8 $6 3 H H n I 0 w v us xgtx3


e

r i y z r8 6  I

This is a completely general formula which permits computation as soon as we specify we assume that is a constant , we have the following:

G$

8 Y6

and we have the usual Black-Scholes formula. When formula

0 0

~ } i{ ~ } i{

8 $6  $6 8 

r 
If is constant, then

g7G 8 Y6 r

~ } !{

8 Y6

x3

)  8 6 
where Similarly,

!{ ~ }

8 Y6

G3

H I

)
G

8 6

8 6

where

330

, . If

CHAPTER 33. Change of num raire e


331 , the value of a European call expiring


Ge

where

This formula also suggests a hedge: at each time , hold bonds.


Ge

We want to verify that this hedge is self-nancing. Suppose we begin with $ and at each time hold shares of stock. We short bonds as necessary to nance this. Will the position in ? If so, the value of the portfolio will always be the bond always be
3 e

and we will have a hedge.

Mathematically, this question takes the following form. Let

At time , hold shares of stock. If is the value of the portfolio at time , then will be invested in the bond, so the number of bonds owned is and the portfolio value evolves according to

The value of the option evolves according to


e 1 1 5A e i 3 Ix3   G$ i IG$ Ge r i 0 '  0 0 V  7e  v 5Ge  e $e v he

If

, will

for

Formulas (3.1) and (3.2) are difcult to compare, so we simplify them by a change of num raire. e This change is justied by the following theorem. Theorem 3.73 Changes of num raire affect portfolio values in the way you would expect. e Proof: Suppose we have a model with assets with prices . At each time , hold shares of asset , , and invest the remaining wealth in asset . Begin with a nonrandom initial wealth , and let be the value of the portfolio at time . The number of shares of asset held at time is

i 3e

  Ge r 5

# 

#%!  "h #

G$

50

 G$

 Ge 5 8 ~ } i r i{ Xu 6  i mey r8 6  i e  e 8 ~ } i r ) i{ Xu 6 e eiy r8 6 v  e ) 0 i WGi  e i r  Wihe 5   0 0  qXq r 


r G$ i 
Ge e

0 i he r 5hWGe  5i

 Gi  5ae

1 1 5

0 i i   WPi i

i Ie

v $e

0 V

i  s AXqq

Gi

x3

Pi

8 $6

As this formula suggests, if at time is

is constant, then for

r 5xi

Ge

shares of stock and short

r 5 0
e i

(3.1)

(3.2)

0   0 i v 0   0   0   Wi 0 i W e 50 W i
Now Therefore,

 0  e 0  e  0  A

S S Q  ' S Q  0 0 v v   S Q` V 0 v S Q` 0 A S  S Q    s $Xq 5

v Q 

 0    0' Q  0 ' A v 


Then

i5 e

e5 0 0 e

e5
Let and 332

 qXq      e
, not

0 ' 0

0 V  A  0 0 i 'v ' A 

   

and we only get to specify Note that

be a num raire, and dene e

evolves according to the equation

, in advance.

CHAPTER 33. Change of num raire e

333

We return to the European call hedging problem (comparison of (3.1) and (3.2)), but we now use the zero-coupon bond as num raire. We still hold e shares of stock at each time . In terms of the new num raire, the asset values are e
i

The portfolio value evolves according to

In the new num raire, the option value formula e


Gi

becomes

This is a homework problem.

  glGi r 5 re i

To show that the hedge works, we must show that

  Gi r 5 Wi i

Ge

 5  $Gi  5  v

Ge

  5Gi  5  v

v 5e

and
Gi

 e

gGe

r 5re i

r 5G$

iGe 

Bond:

he

i Pi

i v e l5i

h  5a e e 0  5ai

Stock:

(3.1)

(3.2)

i  s PqX

Gi

)  5pi

This is the formula for the evolution of a portfolio which holds , and all assets and the portfolio are denominated in units of

shares of asset ,

Ge

 5a

334

@ '

6 v y$

6 a

 1 1 45352 @ '
-maturity bond.

 e

G$

G$

i P

G$

ah


satisfy

v u gs  P  i v u zgs  e P h| i


. The bond prices

G eG$

@ '

 $6 Ph 6
G

6 v |5

G$

34.1 Review of HJM under risk-neutral

Brace-Gatarek-Musiela model

Chapter 34

Forward rate at time for borrowing at time

 z

6 G$6

h h

The interest rate is

where

A simple choice we would like to use is

To implement HJM, you specify a function

where

is the constant volatility of the forward rate. This is not possible because it leads to

volatility of

335

 6 G$ r aIG$ aIG$ 6  6

k W6

 $

aeh

Ix$

|v$G$ P v y$G$ P

e q

i Iq e q i I$
q

x3

$Wi i  i i Iq i

x3

v i   i f
where
s

ge

i f

The problem with the above equation is that the term grows like the square of the forward rate. To see what problem this causes, consider the similar deterministic ordinary differential equation

and Heath, Jarrow and Morton show that solutions to this equation explode before .

This solution explodes at

k Fax3

New variables:

34.2 Brace-Gatarek-Musiela model

336

. We have

Current time

Time to maturity

 i y2

Bond prices:

Forward rates:

(2.4) (2.1) (2.3) (2.2)

$xip|5G v I$ v u g s p   i n p  v u gs   y%h i v 3z 2 C n v u gs p   Y i n v |5G lI$

i a2h

  v u gw gs i y | v u gs $  I | Px$ yv Phx$|v 8$

x$

 e 

@ V @ '

6 i rP

V $

i rx$

i We

v u%h

u%h  v

v u9G

v
$

(2.8)

 6 a 6
$

We will now write

@V   G$ h 6WiP G$ heah @ V   ge $6|5 v ah 6 6 i 6 haP6


rather than
h

We now derive the differentials of

where

CHAPTER 34. Brace-Gatarek-Musiela model

and

differential applies only to rst argument

|%  v h

(2.5),(2.2)

r w @ '  r G d  % e 6 u% v 6 v @ V  e v 6 v $|$ $6 6

I$

Also,

(2.6),(2.4)

v |$ G

6 i WP

v |5he

34.3 LIBOR
(2.1)

I$

Fix time

(say, .

(2.8)

differential applies only to rst argument

year). $ invested at time in a -maturity bond grows to $ 1 at is dened to be the corresponding rate of simple interest:

v y%

 z

|5 v q2 k

, analogous to (2.5) and (2.6) We have

. In this notation, the HJM model is

(2.10) (2.9) (2.7) (2.6) (2.5) 337

@ V

T 5

. Can do this at time by shorting bonds maturing at time . The value of this portfolio at time is one bond maturing at time

We cannot have a log-normal model for because solutions explode as we saw in Section 34.1. For xed positive , we can have a log-normal model for .

a Dgs v u  I | uma v $q

zgs v u  P | $q
Connection with forward rates:

 w gs I v u  i y | $q  | $q  y  $q  E 

gs v u v u w gs i y4$ v vgs u P uv iVB IG$| v uvm


 v gD|F

v $$

v y5

$ i D|vF v v uF|5

d $q " # d " # d $q#  " " d

34.4 Forward LIBOR

338

is still xed. At time , agree to invest $

at time

v yP

yvy5 v k `
r

so

The forward LIBOR

is dened to be the simple (forward) interest rate for this investment:

, with payback of $1 at time

and going long

xed

i y 25

Fgs I$ w v u  i2y5 $ $q ! { $q Fgs w v u

v au%G$

is the continuously compounded rate.

is the simple rate over a period of duration .

34.5 The dynamics of

3'VW

We want to choose

  yv$ Xq  G G (sR6


, appearing in (2.5) so that (4.2) (4.1)

6 h

p @ '  e e |m 6 6 i v v V rIy4$ 6 |F$ |v 6 i v v 6  i v V$W|F| e|v n p  r 6 i v r V$ rIy4 6  v @ ' e V$ 6W|F$ 6 i v v r r i r Gy4 6h v$DPum v  i v n $V$uv  zgu4s v s   | v P | $q $q  zg4 v us    P  %q $q l Fgs w v u i y 
$

$q

@' e V 6Wi|vF 6 v r r  w 6 6 v v r G$ d  i r Gyv4$ d  $$Wiy4| r @'    w $6 $q v% r G d $| $q 6 v     $q | $q


|

 e

@ '

6 v |5

G

r w  6 d  $| v P
Recall (2.9):

. This is the BGM model, and is a subclass of HJM models, for some corresponding to particular choices of .

G W

and

Therefore,

CHAPTER 34. Brace-Gatarek-Musiela model

(4.1), (5.1)
 # 

339

(5.2) (5.1)

 G G (3

 e

@ V

Tv |v  v 6 T v $ $$ r r T

|v $ 7|F$ 6 v 6 v T

 i

@ '

T v $%

 v 6 T v y4 $|$$ I$
Then

 6 i v he WDum 6

V $

uv

 Ve

@ '

v %

  v y4$ 6 e W|F$ 6 V |v  6 i v v $
Therefore,

 i v VWy4$ e|v zg s v u  VWi|vF$ (I  $q i y 




%q

Obtain the initial forward LIBOR curve

from market data. Choose a forward LIBOR volatility function (usually nonrandom)

G (

34.6 Implementation of BGM


5 $T
Take
$

w gs v u

Plugging this into (5.4) yields

Note that (5.3) is equivalent to

But

340

to be given by

(5.4) (5.3) (5.4) (5.3)

v |%h I$|$ ei|hh5iXG T 6 i y4 6 $ T v

6 7

$x

6 i v rIy4$x6

yv

T
Remark 34.3 From (5.3) we have

 e

6 6 

G G

G$5uv  @'  v i 5 iX|G$ v $ eiWG$Th5 @ '  v v i % |5h Ve

i T erhU i T erhUIh5  

i |G

i i i|G

 i eyh aIG$5

 q

uz 1 6 o 1 G yA3

1 G uA3

Because LIBOR gives no rate information on time periods smaller than , we must also choose a partial bond volatility function

for maturities less than from the current time variable .

CHAPTER 34. Brace-Gatarek-Musiela model

With these functions, we can for each

Plugging the solution into (5.3), we obtain

solve (5.4) to obtain

for

is

is o

Remark 34.1 BGM is a special case of HJM with HJMs In BGM, is usually taken to be nonrandom; the resulting

Remark 34.2 (5.4) (equivalently, (5.4)) is a stochastic partial differential equation because of the term. This is not as terrible as it rst appears. Returning to the HJM variables and , set

1 G uz

If we let

and (5.4) and (5.4) become

Then

and we continue recursively.

, then

We saw before (eq. 4.2) that as

and so

generated recursively by (5.3). is random.

. We then solve (5.4) to obtain

(6.1) 341

1 1 A

i W|hd 6

v 5i

 `

G3G G

 F

G3

p  r

G

i 6 hd d

r @ '  Ih i i hd 6

i n

v u gs

 e
Vh

@ ' 

eX

i |h 6

i p S e
G

v %

G$

he

xi

G$5e |G r |1 i T G$5 1 r iX|G$ r X i T


G

G$

G$5|v r euG r q i T

 # " $ d! f " qh # " f $ d # d @ V   Vee v i 5 e h 6

 

Remark 34.4 Although the term in (6.1) has the term to this equation do not explode because

h5
so

Therefore, the limit as

342

of (6.1) is given by equation (2.5):

involving

i 6 |G$a

Let

 y

is a Brownian motion under

Girsanovs Theorem implies that

This is a martingale, and we can use it to switch to the forward measure

 # '& # " $ d!

Q ` 


v u w gs

The solution

34.7 Bond prices

to this stochastic differential equation is given by

From (2.6) we have

. , solutions

h3 G$

he

 v i D|G Rr$h5da yW v i r$h5da i e v |r  r$h5d i i % e


i

yvr5a 1 1 W$h5d i

v yr

yv511A k U

Consider a oating rate interest payment settled in arrears. At time , the oating rate interest the LIBOR at time . A caplet protects its owner by payment due is requiring him to pay only the cap if . Thus, the value of the caplet at time is . We determine its value at times .

v ur

gG

(x

 W|Gd r T i r i r  @ V e Wi|Gd r T  d $ i i rrhdT r v u  @ ' mgs `P Wihd r T  d % i i rGdT I 


h5

a a5
G

i r|Gd r T

r  Ih $ i

@ V

i WrGdT

v u mgs

v u gs

h35

p  rGd r T i

r  Ih $ i

@ V

i WrGdT

h35

aIG$5
so

Vi

@ '

v $

@ $ V e v i u%X|h 6

G$5 e

G$5e rhU i T  i T rhUIG$5

We assume that

and mean

r r  i

and

Case I:

is normal with variance

34.8 Forward LIBOR under more forward measure

CHAPTER 34. Brace-Gatarek-Musiela model

34.9 Pricing an interest rate caplet

From (6.1) we have

is nonrandom. Then

(9.1) (8.2) (8.1) 343

i u 

T Ui

 i 1 1 Xry535A

gGG$5 1 D

r r S S ~i{Ie Q } i5i Q ~ }{  i  i 3Wi e  v 7 v u gs w WxeqB i E d $ Pzd1b x


Set

v u gs  e WxiqB i E  qB

G$5

% v yrhD7e $ x  %
e e

 e

i r v u gs

G$5

i e WG

$  # $ yh3 Gyr v v 5 v v uh8ur


e W5h5d $ i v urG$ i |rG$ v v D|Wh3 he urG$ v e x$G i


5

v  |r(

where is normal under with variance Furthermore, is independent of .

yWG Gi v  e v yrh

From (8.1) and (8.2) we have

v y
.

A
5

Case II: Recall that

1 1 35A

We have

where

344

and mean

r r  i

i |Gh r T

a5h5

and (9.2) is called the Black caplet formula.

b v d urG$

In the case of constant , we have


e

T 7e $

a i $

v |r
e

Then

(9.2) .

gy h

$  $Xqxg    

G3

 Y

 dr T
z

 qq


dr T

X


i F

dr T


x3

x3

 9

dr T

i | r T

 T dU T 
Let us suppose

i W|Gh r T

gGG35

b v d |rh3D7Px $
at time

 e

 e

  s X$a$qXx8a r

a 

Rl

A cap is a series of payments

34.10 Pricing an interest rate cap

CHAPTER 34. Brace-Gatarek-Musiela model

Let

at time

i  `Xq

2 

i rI

The interest rate caplet on

v yr

h3

Then depends on

If we know the caplet price know caplet prices

where (dened in the last section) depends on

34.11 Calibration of BGM

The value at time of the cap is the value of all remaining caplets, i.e.,

is a deterministic function of its second argument, i.e.,

, we can back out the squared volatility

dr T

 XqXx3 $

x3

X $

where

  z zd r T X d r T o o X  qX  $ o

In this case, we may assume that

, we can back out

is constant on each of the intervals has time-zero value . If we (11.1) 345

346 and choose these constants to make the above integrals have the values implied by the caplet prices.

To implement BGM, we need both

, and

We can now solve (or simulate) to get


or equivalently, using the recursive procedure outlined at the start of Section 34.6.

34.12 Long rates

The long rate is determined by long maturity bond prices. Let that is 20 or 30 years. Then be a large xed positive integer, so

where the last equality follows from (4.1). The long rate is

34.13 Pricing a swap


s

 X$|v

 s A$XXq|iyv

|v

Let

be given, and set

7v

Now is the volatility at time of a zero coupon bond maturing at time , it is reasonable to set Since is small (say year), and

 V

dr T

 eD idD$|v  zgs  v u  P $ #  v u gs  I 

i zd

 X

 X

1 G yAX

 $~i}{ $ ~ } i{ % $ yv

1 G uA3

G G UGs

G G `3

G (

sG(Ph T dr T

l $

G a

1 y

$$

IdT $ 6

If we know caplet prices tiate to discover and


for all

, we can back out for all .

and then differen-

hr T

(see (2.6)).

 G$

In contrast to the cap formula, which depends on the term structure model and requires estimation of , the swap formula is generic.


which makes the time

v h5vXqU5  G$ i h  G$

e

|v

i  $G$

X  G Xi hWiXq|i r G Wi  hW hl  i r svqXUv5  hG|v i  G$ 55  G$ G|v i hl v  hGyv qP G e i   i i5x$ d 


for maturity

  hGyv i q   hG|v i e i #  "    h G  hGyv qi e  i  G Q  S e Wi i i  i P h| i 

 m

 i mhe WIx$


i

The swap is the series of payments

CHAPTER 34. Brace-Gatarek-Musiela model

at time

i  RqX

The forward swap rate at time value of the swap equal to zero:
e

 i r5

The value of the swap at time is

Px$

For

1 1 5A

Now

Px$

We compute

yv

so

, the value of the swap is

is the value of

G$

347

348

Chapter 35

Notes and References


35.1 Probability theory and martingales.
Probability theory is usually learned in two stages. In the rst stage, one learns that a discrete random variable has a probability mass function and a continuous random variable has a density. These can be used to compute expectations and variances, and even conditional expectations. Furthermore, one learns how transformations of continuous random variables cause changes in their densities. A well-written book which contains all these things is DeGroot (1986). The second stage of probability theory is measure theoretic. In this stage one views a random variable as a function from a sample space to the set of real numbers . Certain subsets of are called events, and the collection of all events forms a -algebra . Each set in has a probability . This point of view handles both discrete and continuous random variables within the same unifying framework. A conditional expectation is itself a random variable, measurable with respect to the conditioning -algebra. This point of view is indispensible for treating the rather complicated conditional expectations which arise in martingale theory. A well-written book on measure-theoretic probability is Billingsley (1986). A succinct book on measure-theoretic probability and martingales in discrete time is Williams (1991). A more detailed book is Chung (1968). The measure-theoretic view of probability theory was begun by Kolmogorov (1933). The term martingale was apparently rst used by Ville (1939), although the concept dates back to 1934 work of L vy. The rst complete account of martingale theory is Doob (1953). e

35.2 Binomial asset pricing model.


The binomial asset pricing model was developed by Cox, Ross & Rubinstein (1979). Accounts of this model can be found in several places, including Cox & Rubinstein (1985), Dothan (1990) and Ritchken (1987). Many models are rst developed and understood in continuous time, and then binomial versions are developed for purposes of implementation. 349

350

35.3 Brownian motion.


In 1828 Robert Brown observed irregular movement of pollen suspended in water. This motion is now known to be caused by the buffeting of the pollen by water molecules, as explained by Einstein (1905). Bachelier (1900) used Brownian motion (not geometric Brownian motion) as a model of stock prices, even though Brownian motion can take negative values. L vy (1939, 1948) discove ered many of the nonintuitive properties of Brownian motion. The rst mathematically rigorous construction of Brownian motion was carried out by Wiener (1923, 1924). Brownian motion and its properties are presented in a numerous texts, including Billingsley (1986). The development in this course is a summary of that found in Karatzas & Shreve (1991).

35.4 Stochastic integrals.


The integral with respect to Brownian motion was developed by It (1944). It was introduced to o nance by Merton (1969). A mathematical construction of this integral, with a minimum of fuss, is given by ksendal (1995). The quadratic variation of martingales was introduced by Fisk (1966) and developed into the form used in this course by Kunita & Watanabe (1967).

35.5 Stochastic calculus and nancial markets.


Stochastic calculus begins with It (1944). Many nance books, including (in order of increasing o mathematical difculty) Hull (1993), Dothan (1990) and Dufe (1992), include sections on It s o integral and formula. Some other books on dynamic models in nance are Cox & Rubinstein (1985), Huang & Litzenberger (1988), Ingersoll (1987), and Jarrow (1988). An excellent reference for practitioners, now in preprint form, is Musiela & Rutkowski (1996). Some mathematics texts on stochastic calculus are ksendal (1995), Chung & Williams (1983), Protter (1990) and Karatzas & Shreve (1991). Samuelson (1965, 1973) presents the argument that geometric Brownian motion is a good model for stock prices. This is often confused with the efcient market hypothesis, which asserts that all information which can be learned from technical analysis of stock prices is already reected in those prices. According to this hypothesis, past stock prices may be useful to estimate the parameters of the distribution of future returns, but they do not provide information which permits an investor to outperform the market. The mathematical formulation of the efcient market hypothesis is that there is a probability measure under which all discounted stock prices are martingales, a much weaker condition than the claim that stock prices follow a geometric Brownian motion. Some empirical studies supporting the efcient market hypothesis are Kendall (1953), Osborne (1959), Sprenkle (1961), Boness (1964), Alexander (1961) and Fama (1965). The last of these papers discusses other distributions which t stock prices better than geometric Brownian motion. A criticism of the efcient market hypothesis is provided by LeRoy (1989). A provocative article on the source of

CHAPTER 35. Notes and References


stock price movements is Black (1986).

351

The rst derivation of the Black-Scholes formula given in this course, using only It s formula, o is similar to that originally given by Black & Scholes (1973). An important companion paper is Merton (1973), which makes good reading even today. (This and many other papers by Merton are collected in Merton (1990).) Even though geometric Brownian motion is a less than perfect model for stock prices, the Black-Scholes option hedging formula seems not to be very sensitive to deciencies in the model.

35.6 Markov processes.


Markov processes which are solutions to stochastic differential equations are called diffusion processes. A good introduction to this topic, including discussions of the Kolmogorov forward and backward equations, is Chapter 15 of Karlin & Taylor (1981). The other books cited previously, ksendal (1995), Protter (1990), Chung & Williams (1983), and Karatzas & Shreve (1991), all treat this subject. Kloeden & Platen (1992) is a thorough study of the numerical solution of stochastic differential equations. The constant elasticity of variance model for option pricing appears in Cox & Ross (1976). Another alternative model for the stock price underlying options, due to F llmer & Schweizer (1993), has o the geometric Ornstein-Uhlenbeck process as a special case. The Feynman-Kac Theorem, connecting stochastic differential equations to partial differential equations, is due to Feyman (1948) and Kac (1951). A numerical treatment of the partial differential equations arising in nance is contained in Wilmott, Dewynne and Howison (1993, 1995) and also Dufe (1992).

35.7 Girsanovs theorem, the martingale representation theorem, and risk-neutral measures.
Girsanovs Theorem in the generality stated here is due to Girsanov (1960), although the result for constant was established much earlier by Cameron & Martin (1944). The theorem requires a technical condition to ensure that , so that is a probability measure; see Karatzas & Shreve (1991), page 198. The form of the martingale representation theorem presented here is from Kunita & Watanabe (1967). It can also be found in Karatzas & Shreve (1991), page 182. The application of the Girsanov Theorem and the martingale representation theorem to risk-neutral pricing is due to Harrison & Pliska (1981). This methodology frees the Brownian-motion driven model from the assumption of constant interest rate and volatility; these parameters can be random through dependence on the path of the underlying asset, or even through dependence on the paths of other assets. When both the interest rate and volatility of an asset are allowed to be stochastic, the Brownian-motion driven model is mathematically the most general possible for asset prices without jumps.

352 When asset processes have jumps, risk-free hedging is generally not possible. Some works on hedging and/or optimization in models which allow for jumps are Aase (1993), Back (1991), Bates (1988,1992), Beinert & Trautman (1991), Elliott & Kopp (1990), Jarrow & Madan (1991b,c), Jones (1984), Madan & Seneta (1990), Madan & Milne (1991), Mercurio & Runggaldier (1993), Merton (1976), Naik & Lee (1990), Schweizer (1992a,b), Shirakawa (1990,1991) and Xue (1992). The Fundamental Theorem of Asset Pricing, as stated here, can be found in Harrison & Pliska (1981, 1983). It is tempting to believe the converse of Part I, i.e., that the absence of arbitrage implies the existence of a risk-neutral measure. This is true in discrete-time models, but in continuous-time models, a slightly stronger condition is needed to guarantee existence of a risk-neutral measure. For the continuous-time case, results have been obtained by many authors, including Stricker (1990), Delbaen (1992), Lakner (1993), Delbaen & Schachermayer (1994a,b), and Fritelli & Lakner (1994, 1995). In addition to the fundamental papers of Harrison & Kreps (1979), and Harrison & Pliska (1981, 1983), some other works on the relationship between market completeness and uniqueness of the risk-neutral measure are Artzner & Heath (1990), Delbaen (1992), Jacka (1992), Jarrow & Madan (1991a), M ller (1989) and Taqqu & Willinger (1987). u

35.8 Exotic options.


The reection principle, adjusted to account for drift, is taken from Karatzas & Shreve (1991), pages 196197. Explicit formulas for the prices of barrier options have been obtained by Rubinstein & Reiner (1991) and Kunitomo & Ikeda (1992). Lookback options have been studied by Goldman, Sosin & Gatto (1979), Goldman, Sosin & Shepp (1979) and Conz & Viswanathan (1991). e Because it is difcult to obtain explicit formulas for the prices of Asian options, most work has been devoted to approximations. We do not provide an explicit pricing formula here, although the partial differential equation given here by the Feynman-Kac Theorem characterizes the exact price. Bouaziz, Bryis & Crouhy (1994) provide an approximate pricing formula, Rogers & Shi (1995) provide a lower bound, and Geman & Yor (1993) obtain the Laplace transform of the price.

35.9 American options.


A general arbitrage-based theory for the pricing of American contingent claims and options begins with the articles of Bensoussan (1984) and Karatzas (1988); see Myneni (1992) for a survey and additional references. The perpetual American put problem was solved by McKean (1965). Approximation and/or numerical solutions for the American option problem have been proposed by several authors, including Black (1975), Brennan & Schwartz (1977) (see Jaillet et al. (1990) for a treatment of the American option optimal stopping problem via variational inequalities, which leads to a justication of the Brennan-Schwartz algorithm), by Cox, Ross & Rubinstein (1979) (see Lamberton (1993) for the convergence of the associated binomial and/or nite difference schemes)

CHAPTER 35. Notes and References

353

and by Parkinson (1977), Johnson (1983), Geske & Johnson (1984), MacMillan (1986), Omberg (1987), Barone-Adesi & Whalley (1987), Barone-Adesi & Elliott (1991), Bunch & Johnson (1992), Broadie & Detemple (1994), and Carr & Faguet (1994).

35.10 Forward and futures contracts.


The distinction between futures contracts and daily resettled forward contracts has only recently been recognized (see Margrabe (1976), Black (1976)) and even more recently understood. Cox, Ingersoll & Ross (1981) and Jarrow & Oldeld (1981) provide a discrete-time arbitrage-based analysis of the relationship between forwards and futures, whereas Richard & Sundaresan (1981) study these claims in a continuous-time, equilibrium setting. Our presentation of this material is similar to that of Dufe & Stanton (1992), which also considers options on futures, and to Chapte 7 of Dufe (1992). For additional reading on forward and futures contracts, one may consult Dufe (1989).

35.11 Term structure models.


The Hull & White (1990) model is a generalization of the constant-coefcent Vasicek (1977) model. Implementations of the model appear in Hull & White (1994a,b). The Cox-Ingersoll-Ross model is presented in (1985a,b). The presentations of these given models here is taken from Rogers (1995). Other surveys of term structure models are Dufe & Kan (1994) and Vetzal (1994). A partial list of other term structure models is Black, Derman & Toy (1990), Brace & Musiela (1994a,b), Brennan & Schwartz (1979, 1982) (but see Hogan (1993) for discussion of a problem with this model), Dufe & Kan (1993), Ho & Lee (1986), Jamshidian (1990), and Longstaff & Schwartz (1992a,b). The continuous-time Heath-Jarrow-Morton model appears in Heath, Jarrow & Morton (1992), and a discrete-time version is provided by Heath, Jarrow & Morton (1990). Carverhill & Pang (1995) discuss implementation. The Brace-Gatarek-Musiela variation of the HJM model is taken from Brace, et al. (1995). A summary of this model appears as Reed (1995). Related works on term structure models and swaps are Flesaker & Hughston (1995) and Jamshidian (1996).

35.12 Change of num raire. e


This material in this course is taken from Geman, El Karoui and Rochet (1995). Similar ideas were used by by Jamshidian (1989). The Merton option pricing formula appears in Merton (1973).

35.13 Foreign exchange models.


Foreign exchange options were priced by Biger & Hull (1983) and Garman & Kohlhagen (1983). The prices for differential swaps have been worked out by Jamshidian (1993a, 1993b) and Brace & Musiela (1994a).

354

35.14 REFERENCES

(1993) Aase, K. K., A jump/diffusion consumption-based capital asset pricing model and the equity premium puzzle, Math. Finance 3(2), 6584. (1961) Alexander, S. S., Price movements in speculative markets: trends or random walks, Industrial Management Review 2, 726. Reprinted in Cootner (1964), 199218. (1990) Artzner, P. & Heath, D., Completeness and non-unique pricing, preprint, School of Operations Research and Industrial Engineering, Cornell University. (1900) Bachelier, L., Th orie de la sp culation, Ann. Sci. Ecole Norm. Sup. 17, 2186. Reprinted e e in Cootner (1964). (1991) Back, K., Asset pricing for general processes, J. Math. Econonics 20, 371395. (1991) Barone-Adesi, G. & Elliott, R., Approximations for the values of American options, Stoch. Anal. Appl. 9, 115131. (1987) Barone-Adesi, G. & Whalley, R., Efcient analytic approximation of American option values, J. Finance 42, 301320. (1988) Bates, D. S., Pricing options under jump-diffusion processes, preprint, Wharton School of Business, University of Pennsylvania. (1992) Bates, D. S., Jumps and stochastic volatility: exchange rate processes implicit in foreign currency options, preprint, Wharton School of Business, University of Pennsylvania. (1991) Beinert, M. & Trautmann, S., Jump-diffusion models of stock returns: a statistical investigation, Statistische Hefte 32, 269280. (1984) Bensoussan, A., On the theory of option pricing, Acta Appl. Math. 2, 139158. (1983) Biger, H. & Hull, J., The valuation of currency options, Financial Management, Spring, 2428. (1986) Billingsley, P., Probability and Measure, 2nd edition, Wiley, New York. (1975) Black, F., Fact and fantasy in the use of options, Financial Analysts J., 31, 3641, 6172.

CHAPTER 35. Notes and References


(1976) Black, F., The pricing of commodity contracts, J. Financial Economics 3. (1986) Black, F., Noise, J. Finance 41, 529543.

355

(1990) Black, F., Derman, E. & Toy, W., A one-factor model of interest rates and its application to treasury bond options, Fin. Anal. Journal, 3339. (1973) Black, F. & Scholes, M., The pricing of options and corporate liabilities, J. Political Economy 81, 637659. (1964) Boness, Some evidence of the protability of trading put and call options. In Cootner (1964), 475496. (1994) Bouaziz, L., Bryis, E. & Crouhy, M., The pricing of forward-starting Asian options, J. Banking Finance. (1994a) Brace, A. & Musiela, M., Swap derivatives in a Gaussian HJM framework, preprint, School of Mathematics, University of New South Wales. (1994b) Brace, A. & Musiela, M., A multi-factor Gauss Markov implementation of Heath, Jarrow and Morton, Math. Finance 4, 259283. (1995) Brace, A.. Gatarek, D. & Musiela, M., The market model of interest rate dynamics, preprint, School of Mathematics, University of New South Wales. (1977) Brennan, M. J. & Schwartz, E. S., The valuation of the American put option, J. Finance 32, 449462. (1979) Brennan, M. J. & Schwartz, E. S., A continuous-time approach to the pricing of bonds, J. Banking Finance 3, 133155. (1982) Brennan, M. J. & Schwartz, E. S., An equilibrium model of bond pricing and a test of market efciency, J. Financial Quantitative Analysis 17, 301329. (1994) Broadie, M. & Detemple, J., American option values: new bounds, approximations and a comparison of existing methods, preprint, Graduate School of Business, Columbia University. (1992) Bunch, D. & Johnson, H., A simple and numerically efcient valuation method for American puts, using a modied Geske-Johnson approach, J. Finance 47, 809816. (1944) Cameron, R. H. & Martin, W. T., Transformation of Wiener integrals under translations,

356 Ann. Math. 45, 386396. (1994) Carr, P. & Faguet, D., Fast accurate valuation of American options, preprint, Johnson School of Business, Cornell University. (1995) Carverhill, A. & Pang, K., Efcient and exible bond option valuation in the Heath, Jarrow, and Morton framework, J. Fixed Income, September, 7077. (1968) Chung, K. L. A Course in Probability Theory, Academic Press, Orlando, FL. (1983) Chung, K. L. & Williams, R. J. (1983) Introduction to Stochastic Integration, Birkh user, a Boston. (1991) Conz , A. & Viswanathan, R., Path dependent options: the case of lookback options, J. e Finance 46, 18931907. (1964) Cootner, P. H. (editor), The Random Character of Stock Market Prices, MIT Press, Cambridge, MA. (1981) Cox, J. C., Ingersoll, J. E., & Ross, S, The relation between forward prices and futures prices, J. Financial Economics 9, 321346. (1985a) Cox, J. C., Ingersoll, J. E. & Ross, S., An intertemporal general equilibrium model of asset prices, Econometrica 53, 363384. (1985b) Cox, J. C., Ingersoll, J. E. & Ross, S., A theory of the term structure of interest rates, Econometrica 53, 373384. (1976) Cox, J. C. & Ross, S. A., The valuation of options for alternative stochastic processes, J. Fin. Economics 3, 145166. (1979) Cox, J. C., Ross, S. A. & Rubinstein, M., Option pricing: a simplied approach, J. Financial Economics 7, 229263. (1985) Cox, J. C. & Rubinstein, M. Options Markets, Prentice Hall, Englewood Cliffs, NJ. (1986) DeGroot, M., Probability and Statistics, 2nd edition, Addison Wesley, Reading, MA. (1992) Delbaen, F., Representing martingale measures when asset prices are continuous and bounded, Math. Finance 2(2), 107130.

CHAPTER 35. Notes and References

357

(1994a) Delbaen, F. & Schachermayer, W., A general version of the fundamental theorem of asset pricing, Mathematische Annalen 300, 463520. (1994b) Delbaen, F. & Schachermayer, W., Arbitrage and free-lunch with bounded risk, for unbounded continuous processes, Math. Finance 4(4), 343348. (1953) Doob, J., Stochastic Processes, Wiley, New York. (1990) Dothan, M. U., Prices in Financial Markets, Oxford University Press, New York. (1989) Dufe, D., Futures Markets, Prentice-Hall, Englewood Cliffs, NJ. (1992) Dufe, D., Dynamic Asset Pricing Theory, Princeton University Press, Princeton. (1993) Dufe, D. & Kan, R., A yield-factor model of interest rates, preprint. (1994) Dufe, D. & Kan, R., Multi-factor term structure models, Phil. Transactions Royal Society London A 347, 577586. (1992) Dufe, D. & Stanton, R., Pricing continuously resettled contingent claims, J. Economic Dynamics and Control, 16, 561573. (1905) Einstein, A. On the movement of small particles suspended in a stationary liquid demanded by the molecular-kinetic theory of heat, Ann. Physik 17. (1990) Elliott, R. J. & Kopp, P. E., Option pricing and hedge portfolios for Poisson processes, Stoch. Anal. Appl. 9, 429444. (1965) Fama, E. The behavior of stock-market prices, J. Business 38, 34104. (1948) Feynman, R., Space-time approach to nonrelativistic quantum mechanics, Rev. Mod. Phys. 20, 367387. (1966) Fisk, D. L., Sample quadratic variation of continuous, second-order martingales, Z. Wahrscheinlichkeitstheorie verw. Gebiete 6, 273278. (1995) Flesaker, B. & Hughston, L., Positive interest, preprint. (1993) F llmer, H. & Schweizer, M., A microeconomic approach to diffusion models for stock o prices, Math. Finance 3, 123.

358 (1994) Fritelli, M. & Lakner, P., Almost sure characterization of martingales, Stochastics 49, 181 190. (1995) Fritelli, M. & Lakner, P., Arbitrage and free-lunch in a general nance market model: the fundamental theorem of asset pricing, IMA Vol 65: Mathematical Finance, 8992, (M. Davis, D. Dufe, W. Fleming & S. Shreve, eds.), Springer-Verlag, New York. (1983) Garman, M. & Kohlhagen, S., Foreign currency option values, J. International Money Finance 2, 231237. (1995) Geman, H., El Karoui, N. & Rochet, J.-C., Changes of num raire, changes of probability e measure, and option pricing, J. Applied Probability 32, 443458. (1993) Geman, H. & Yor, M., Bessel processes, Asian options, and perpetuities, Mathematical Finance 3, 349375. (1984) Geske, R. & Johnson, H., The American put-option valued analytically, J. Finance 39, 1511 1524. (1960) Girsanov, I. V., On transforming a certain class of stochastic processes by absolutely continuous substitution of measures, Theory Probability Applications 5, 285301. (1979) Goldman, B., Sosin, H. & Gatto, M., Path dependent options: Buy at the low, sell at the high, J. Finance 34, 6984. (1979) Goldman, B., Sosin, H. & Shepp, L. A., On contingent claims that insure ex-post optimal stock market timing, J. Finance 34, 401414. (1979) Harrison, J. M. & Kreps, D. M., Martingales and arbitrage in multiperiod securities markets, J. Economic Theory 20, 381408. (1981) Harrison, J. M. & Pliska, S. R., Martingales and stochastic integrals in the theory of continuous trading, Stochastic Processes Applications 11, 215260. (1983) Harrison, J. M. & Pliska, S. R., A stochastic calculus model of continuous trading: complete markets, Stochastic Processes Applications 15, 313316. (1990) Heath, D., Jarrow, R. & Morton, A., Bond pricing and the term structure of interest rates: a discrete time approximation, J. Financial Quantitative Analysis 25, 419440. (1992) Heath, D., Jarrow, R. & Morton, A., Bond pricing and the term structure of interest rates: a

CHAPTER 35. Notes and References


new methodology, Econometrica 60, 77105.

359

(1993) Hogan, M., Problems in certain two-factor term structure models, Annals Applied Probability 3, 576581. (1986) Ho, T. S. Y. & Lee, S.-B., Term structure movements and pricing interest rate contingent claims, J. Finance 41, 10111029. (1988) Huang, C.-f. & Litzenberger, R., Foundations for Financial Economics, North Holland, Amsterdam. (1993) Hull, J., Options, Futures, and other Derivative Securities, Prentice Hall, Englewood Cliffs, NJ. (1990) Hull, J. & White, A., Pricing interest rate derivative securities, Rev. Financial Studies 3, 573592. (1994a) Hull, J. & White, A., Numerical procedures for implementing term structure models I: single-factor models, J. Derivatives, Fall, 716. (1994b) Hull, J. & White, A., Numerical procedures for implementing term structure models II: two factor models, J. Derviatives, Winter. (1987) Ingersoll, J. E., Theory of Financial Decision Making, Rowman & Littleeld. (1944) It , K., Stochastic integral, Proc. Imperial Acad. Tokyo 20, 519524. o (1992) Jacka, S., A martingale representation result and an application to incomplete nancial markets, Math. Finance 2, 239250. (1990) Jaillet, P., Lamberton, D. & Lapeyre, B., Variational inequaltiies and the pricing of American options, Acta Appl. Math. 21, 263289. (1989) Jamshidian, F., An exact bond option pricing formula, J. Finance 44, 205209. (1990) Jamshidian, F., The preference-free determination of bond and option prices from the spot interest rate, Advances Futures & Options Research 4, 5167. (1993a) Jamshidian, F., Price differentials, Risk Magazine 6/7.

360 (1993b) Jamshidian, F., Corralling quantos, Risk Magazine 7/3. (1996) Jamshidian, F., Sorting out swaptions, Risk Magazine 9/3, 5960. (1988) Jarrow, R., Finance Theory, Prentice Hall, Englewood Cliffs, NJ. (1991a) Jarrow, R. & Madan, D., A characterization of complete security markets on a Brownian ltration, Math. Finance 1(3), 3143. (1991b) Jarrow, R. & Madan, D., Valuing and hedging contingent claims on semimartingales, preprint, Johnson School of Business, Cornell University. (1991c) Jarrow, R. & Madan, D., Option pricing using the term structure of interest rates to hedge systematic discontinuities in asset returns, preprint, Johnson School of Business, Cornell University. (1981) Jarrow, R. & Oldeld, G., Forward contracts and futures contracts, J. Financial Economics 9, 373382. (1983) Johnson, H., An analytic approximation for the American put price, J. Financial Quant. Analysis 18, 141148. (1984) Jones, E. P., Option arbitrage and strategy with large price changes, J. Financial Economics 13, 91113. (1951) Kac, M., On some connections between probability theory and differential and integral equations, Proc. 2nd Berkeley Symposium on Math. Stat. & Probability, 189215. (1988) Karatzas, I., On the pricing of American options, App. Math. Optimization 17, 3760. (1991) Karatzas, I. & Shreve, S. E., Brownian Motion and Stochastic Calculus, Springer-Verlag, New York. (1981) Karlin, S. & Taylor, H. M., A Second Course in Stochastic Processes, Academic Press, Orlando, FL. (1953) Kendall, M. G., Analysis of economic time-series Part I: prices, J. Royal Statistical Society 96, Part I, 1125. Reprinted in Cootner (1964), 8599. (1992) Kloeden, P. E. & Platen, E., Numerical Solution of Stochastic Differential Equations, SpringerVerlag, New York.

CHAPTER 35. Notes and References

361

(1933) Kolmogorov, A. N., Grundbegriffe der Wahrscheinlichkeitsrechnung, Ergebnisse Mathematik 2. English translation: Foundations of Probability Theory, Chelsea, New York, 1950. (1967) Kunita, H. & Watanabe, S., On square-integrable martingales, Nagoya Math. Journal 30, 209245. (1992) Kunitomo, N. & Ikeda, M., Pricing options with curved boundaries, Mathematical Finance 2, 275298. (1993) Lakner, P., Martingale measures for a class of right-continuous processes, Math. Finance 3, 4353. (1993) Lamberton, D., Convergence of the critical price in the approximation of American options, Math. Finance 3, 179190. (1989) LeRoy, S. F., Efcient capital markets and martingales, J. Economic Literature 27, 1583 1621. (1939) L vy, P., Sur certains processus stochastiques homog nes, Compositio Math. 7, 283339. e e (1948) L vy, P., Processus Stochastiques et Mouvement Brownian, Gauthier-Villars, Paris. e (1992a) Longstaff, F. A. & Schwartz, E. S., Interest rate volatility and the term structure: a twofactor general equilibrium model, J. Finance 47, 12591282. (1992b) Longstaff, F. A. & Schwartz, E. S., A two-factor interest rate model and contingent claim valuation, J. Fixed Income 3, 1623. (1986) MacMillan, L., Analytic approximation for the American put option, Adv. Futures Options Research 1, 119139. (1990) Madan, D. & Seneta, E., The V. G. model for share returns, J. Business 63, 511524. (1991) Madan, D. & Milne, F., Option pricing with V. G. martingale components, Math. Finance 1(4), 3956. (1976) Margrabe, W., A theory of forward and futures prices, preprint, Wharton School of Business, University of Pennsylvania. (1965) McKean, H. P., Jr., Appendix to P. A. Samuelson (1965): A free boundary problem for the heat equation arising from a problem in mathematical economics, Industrial Management Review

362 6, 3239. (1993) Mercurio, F. & Runggaldier, W., Option pricing for jump diffusions: approximations and their interpretation, Math. Finance 3, 191200. (1969) Merton, R. C., Lifetime portfolio selection under uncertainty: the continuous-time case, Review Economics Statistics 51, 247257. (1973) Merton, R. C., Theory of rational option pricing, Bell J. Econ. Management Science 4, 141 183. (1976) Merton, R. C., Option pricing when underlying stock returns are discontinuous, J. Fin. Economics 3, 125144. (1990) Merton, R. C., Continuous-Time Finance, Basil Blackwell, Cambridge, 1990. (1989) M ller, S. On complete securities markets and the martingale property on securities prices, u Economic Letters 31 3741. (1996) Musiela, M. & Rutkowski, M., Arbitrage Pricing of Derivative Securities. Theory and Applications, Springer-Verlag, New York, to appear. (1992) Myneni, R., The pricing of the American option, Ann. Applied Probability 2, 123. (1990) Naik, V. & Lee, M., General equilibrium pricing of options on the market portfolio with discontinuous returns, Rev. Financial Studies 3, 493521. (1995) ksendal, B., Stochastic Differential Equations, 4th Edition, Springer-Verlag, New York. (1987) Omberg, E., The valuation of American puts with exponential exercise policies, Adv. Futures Option Research 2, 117142. (1959) Osborne, M. F. M., Brownian motion in the stock market, Operations Research 7, 145173. Reprinted in Cootner (1964), 100128. (1977) Parkinson, M., Option pricing: the American put, J. Business 50, 2136. (1990) Protter, P., Stochastic Integration and Differential Equations, Springer-Verlag, New York.

 qX

(1995) Reed, N., If the cap ts

, Risk 8(8), 3435.

CHAPTER 35. Notes and References

363

(1981) Richard, S. F. & Sundaresan, M., A continuous-time equilibrium model of forward and futures prices in a multigood economy, J. Financial Economics, 9, 347371. (1987) Ritchken, P. Options: Theory, Strategy and Applications, Harper Collins. (1995) Rogers, L. C. G., Which model of term-structure of interest rates should one use?, IMA Vol. 65: Mathematical Finance, 93116, (M. Davis, D. Dufe, W. Fleming & S. Shreve, editors), Springer-Verlag, New York. (1995) Rogers, L. C. G. & Shi, Z., The value of an Asian option, J. Applied Probability, to appear. (1991) Rubinstein, M. & Reiner, E., Breaking down the barriers, Risk Magazine 4/8, 2835. (1965) Samuelson, P. A., Proof that properly anticipated prices uctuate randomly, Industrial Management Review 4, 4150. (1973) Samuelson, P. A., Mathematics of speculative prices, SIAM Review 15, 142. (1992a) Schweizer, M., Martingale densities for general asset prices, J. Math. Economics 21, 123 131. (1992b) Schweizer, M., Mean-variance hedging for general claims, Annals Appl. Probab. 2, 171 179. (1990) Shirakawa, H., Optimal dividend and portfolio decisions with Poisson and diffusion-type return processes, preprint IHSS 90-20, Tokyo Institute of Technology. (1991) Shirakawa, H., Interest rate option pricing with Poisson-Gaussian forward rate curve processes, Math. Finance 1(4), 7794. (1961) Sprenkle, C. M., Warrant prices as indicators of expectations and preferences, Yale Economic Essays 1, 178231. Reprinted in Cootner (1964), 412474. (1990) Stricker, C., Arbitrage et lois de martingales, Ann. Inst. henri Poincar 26, 451460. e (1987) Taqqu, M. & Willinger, W., The analysis of nite security markets using martingales, Adv. Appl. Probab. 19, 125. (1977) Vasicek, O., An equilibrium characterization of the term structure, J. Financial Economics 5, 177188.

364 (1994) Vetzal, K. R., A survey of stochastic continuous time models of the term structure of interest rates, Insurance: Mathematics and Economics 14, 139161. (1939) Ville, J., Etude Critique de la Notion du Collectif, Gauthier-Villars, Paris. (1923) Wiener, N., Differential spaces, J. Math. Physics 2, 131174. (1924) Wiener, N., Un probl me de probabilit es d nombrables, Bull. Soc. Math. France 52, 569 e e 578. (1991) Williams, D., Probability with Martingales, Cambridge University Press, Cambridge. (1993) Wilmott, P., Dewynne, J. & Howison, S., Option Pricing, Oxford Financial Press, PO Box 348, Oxford, OX4 1DR, UK. (1995) Wilmott, P., Dewynne, J. & Howison, S., The Mathematics of Financial Derivatives, Cambridge University Press, Cambridge. (1992) Xue, X. X., Martingale representation for a class of processes with independent increments, and its applications, in Lecture Notes in Control and Information Sciences 177, 279311, SpringerVerlag, Berlin.