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 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Definition 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-financing 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 Jensen’s Inequality 7.1 7.2 7.3 Jensen’s Inequality for Conditional Expectations . . . . . . . . . . . . . . . . . . . Optimal Exercise of an American Call . . . . . . . . . . . . . . . . . . . . . . . . Stopped Martingales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

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

¢

3

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 Reflection 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 finite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

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 Itˆ 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 Itˆ Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 o 14.6 Itˆ integral of an elementary integrand . . . . . . . . . . . . . . . . . . . . . . . . 158 o 14.7 Properties of the Itˆ integral of an elementary process . . . . . . . . . . . . . . . . 159 o 14.8 Itˆ integral of a general integrand . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 o

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

15.1 Itˆ ’s formula for one Brownian motion . . . . . . . . . . . . . . . . . . . . . . . . 167 o 15.2 Derivation of Itˆ ’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 Itˆ 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 Girsanov’s 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 Reflection 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: first 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

. . . . . . . . . . . . . . . . . . . . . . . . . . . . 269 27. . . . . 277 28. . . . . . . . . . . . . . . . . . . . . . . . .1 American option with convex payoff function . . . . . . . . . . . . .3 Terminology . . 280 28. . . . . . . . . . . . . . . . 272 27.8 Implementation of the Heath-Jarrow-Morton model . . . . . . . . . . 270 27. . . .4 Forward rate agreement . . . .5 Recovering the interest from the forward rate . . . . . . . . . . . 252 25. . .2 Dividend paying stock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276 28. . 261 26 Options on dividend-paying stocks 263 26. . . . . . . . . . . . . . . . . . . . . . . . . . . . 269 27. . . . . . . . . . . . . . . . . . . . 276 28. . . . . . . . . . . . . . . . . . . . . . . . . 281 29 Gaussian processes 285 28. . . 264 26. . . . . .9 Perpetual American contingent claim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259 25. . . . . . . . . . . . 257 25. .2 Some interest-rate dependent assets . . . . .6 Computing arbitrage-free bond prices: Heath-Jarrow-Morton method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Hedging at time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7 Value of the perpetual American put . . . . . . 266 267 27 Bonds. . . . . . . 279 28. . 259 25.1 An example: Brownian Motion . . . . . . . . . . . . . . . . . . . . .4 Cash flow from a future contract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . forward contracts and futures 27. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8 Hedging the put . . . . . . . . . . . . . . . . . . . . . . . . 273 28 Term-structure models 275 28. .7 25. . . . . . . . . . . . . . . 260 25. . . . . . . .1 Forward contracts . . . . . . . . .7 Checking for absence of arbitrage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286 30 Hull and White model 293 ' $ (&%# ! " . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10Perpetual American call . . . . . . . . 263 26. . . . . . . . . . . . . 256 25. . . . . . . . . . . . . . .12American contingent claim with expiration . . . . . . . . .1 Computing arbitrage-free bond prices: first method .3 Future contracts . . . .6 Backwardation and contango . . . . .2 Hedging a forward contract . . . . . . . 277 28. . . . . 272 27. . . . 278 29. .5 Forward-future spread . . . . . .11Put with expiration . . . . . . .6 Perpetual American put . . .

. . . . . . . . . . . . . . . . . . . . . . . 343 34. . . . . .5 Option on a bond . . . . . . . .3 Calibration of the Hull & White model . . . . . . . . . . . . . . . . . . . . . . 313 31. . . . . . .1 Bond price as num´ raire . . . . . . . . . . . . . . . .1 Equilibrium distribution of ' 6$3 75542 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Merton option pricing formula . . . . . . . . . . . 297 30. . . . . . . . 306 31. . . . . . . . . . . 328 e 33. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Fiddling with the formulas . . . . . . . . . . . . . . . . . . . . . . . .8 30. .4 Forward LIBOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321 32. . . 313 31. . . . 329 34 Brace-Gatarek-Musiela model 335 . . . . . . . . . . . . . . . . . . .3 Cox-Ingersoll-Ross equilibrium density . . . . . . . . . . . 296 30. . . . . . . . . . 337 34. . . . . . . . . . . . . . . . . . . . .5 The dynamics of . . . . . . . . . . 323 33 Change of num´ raire e 325 33.8 Forward LIBOR under more forward measure . . . . . . . . . 316 319 32 A two-factor model (Duffie & Kan) 32. . . . 336 34. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Dynamics of the bond price . . . . . . . . . .4 Bond prices in the CIR model .6 Deterministic time change of CIR model . . . . . . . . . . . . 327 e 33. . . . . . . . . . . . . . . . . . .1 Review of HJM under risk-neutral ' £A $ )BB&@9 8 32. . .3 Calibration . 340 34. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 LIBOR . . . . . . . . 299 31 Cox-Ingersoll-Ross model 303 . . . . . . . . . . . . . . . 338 34. . . . . . . . . . . . . . 342 34. . . . . . . . . . . . .6 Implementation of BGM . . . . . . . . . . . . . . . . .7 Bond prices . . . . . . . . . . . . . . . 295 30. . .8 Tracking down in the time change of the CIR model . . . . . . . . 310 31. . . . . . . . . . . . . . 306 31. . . . . . . . . . . . . . .2 Brace-Gatarek-Musiela model . . . . 338 ¥¦ 34. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Stock price as num´ raire . . . . . . . 320 ' $ 10)# 31. . . 315 31. . . . . . . .1 Non-negativity of . . . . . .4 Option on a bond . . . . . . . . . . . 309 31. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Zero-coupon bond prices . . . . . . . . . . . . 335 34. . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Kolmogorov forward equation . . . . . . . . . .

. . . . . . . .11Term structure models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10Pricing an interest rate cap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349 35. . . . . 354 . . . . . . . . . . . . . . . 350 35. . 349 35. . . . . . . . . . . . . . . . 352 35. . 352 35. . . . 346 34. . . . . . . . . . and risk-neutral measures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Probability theory and martingales. . . . . . . . . . . . . . . .6 Markov processes. . . . . . . . . . . . . . . . .5 Stochastic calculus and financial markets. . . . . . . . . . . . . . . . . . . . . . . . . 345 34. . . . . . . . . . . . . .13Foreign exchange models. .13Pricing a swap . 350 35. . . . . . . . . . . . . . . . the martingale representation theorem. . . . . . .351 35. . . .9 American options. . . . . . . . . . . . . . 351 35. . . . . . . . . . . . . . . . . . . . . . .12Long rates . . . . . . . . . . . . . . . . . . . . . 353 35. . . . . . . . . . . . . . . . . . .8 Exotic options. . . . . . . .4 Stochastic integrals. . . . . . . . . . . . . . . . . .9 34. . . . . . . . . . . . . 353 e 35. . . . . . . . . . . . . . . . . .9 Pricing an interest rate caplet . . . . . . . . . . . . . . 346 35 Notes and References 349 35. . . . . . . . . . . . . . . . . . . 345 34. . . . . . 353 35. . . . . 353 35. . . . . . . . . . .10Forward and futures contracts. . . . . . . . . . . . . . . . .12Change of num´ raire. . . . . . . . . . . 350 35. . . .2 Binomial asset pricing model. 343 34. . . . . . . . . . . . . . . . . . . . .7 Girsanov’s theorem. . . . . . . . . . . . . . . . . . . .14REFERENCES . . . . . . . . . .3 Brownian motion. . . . . . . . . . . . . . . . .11Calibration of BGM . . .

10 .

and when we get a “Head. assuming that at each step. It is common to also have . In this course. and change of the stock price from to represents an upward movement. and this will be the case in many of our examples. for what we are about to do we need to assume only (1. the stock price will change to one of two possible values. With this third motivation in mind. we shall use it for both these purposes.Chapter 1 Introduction to Probability Theory 1. but when we get a “Tail.1) . Of course.” the stock price moves up. stock price movements are much more complicated than indicated by the binomial asset pricing model. within this model the concept of arbitrage pricing and its relation to risk-neutral pricing is clearly illuminated. There are two positive numbers. Let us imagine that we are tossing a coin. strictly speaking. Secondly. with such that at the next period.2) below. In the binomial asset pricing model. the stock price will be either or . we develop notation for the binomial model which is a bit different from that normally found in practice. Typically. Thirdly. we take and to satisfy . and by if it 11 F 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 © F § D EC (1.1) and (1. the model is used in practice because with a sufficient number of steps. so change of the stock price from to represents a downward movement.” the price moves down. We consider this simple model for three reasons. and . First of all. within the binomial model we can develop the theory of conditional expectations and martingales which lies at the heart of continuous-time models.1 The Binomial Asset Pricing Model The binomial asset pricing model provides a powerful tool to understand arbitrage pricing theory and probability theory. computationally tractable approximation to continuous-time models. However. it provides a good. We denote the price at time by if the toss results in head (H). Let us begin with an initial positive stock price . we model stock prices in discrete time.

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.1. We denote the -th component of by .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). Thus. . „  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@€1€w@1wd€1@€‰Bwˆw"€wdBwd‡u† „ ‚ D §  ' e$ ! C §  ' e e$ q A D C § F  ' e$ ! C F  ' c e$ q 7EC q t@0s4Qt@€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 i™reAuqx’1l!–d’$k’  p  f vWF g D C !q j§ '™ ’Aq ’A! $  sresxeihd’g•’ q fC c e c  @w•’ ’ '# n ©$ uHh1Sm ! xC ’ “ ’ ” t’ ƒ ”C ’ DC . and the elements of are called sample points. For example. we often write . 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. the price will be one of: After three tosses. For the moment. Each sample point represents a path through the tree. depends on only the first component of . at time depends on the coin tosses. 1. for while depends on all of . when . In this case. Sometimes we will use notation such just to record more explicitly how depends on . let us assume that the third toss is the last one and denote by the set of all possible outcomes of the three tosses. Example 1. We have then the binomial “tree” of possible stock prices shown in Fig.1 Set .1: Binomial tree of stock prices with . although not all of them result in different stock prices at time . this notation does not quite tell the whole story. we have . there are eight possible coin sequences. depends on only the first two components of . To emphasize this. The set of all possible outcomes of a random experiment is called the sample space for the experiment. and does not depend on at all. and . and . we introduce a money market with interest rate . After the second toss. The stock price Actually. each sample point is a sequence of length three. To complete our binomial asset pricing model. $1 invested in the money market becomes in the next period.

except maybe once upon a time in Switzerland) or .CHAPTER 1. because in a many applications of the model. In the latter case. The inequality cannot happen unless either is negative (which never happens. where this quantity might be negative. if . You now have an obligation to pay off if and to pay off if . in such an application. where is still to be determined. In either case. You will also own shares of stock. Our first task is to compute the arbitrage price of this option at time zero. You can use position in the option by buying the proceeds of the sale of the option for this purpose. and so is worth at time if is positive and is otherwise worth zero. Introduction to Probability Theory 13 rate for both borrowing and lending. Thus. it just goes up less than if we had gotten a head. you want to choose and so that (1. If is more than necessary to buy the shares of stock. At the time you sell the option. One should borrow money at interest rate and invest in the stock. With the stock as the underlying asset. If the stock goes up. 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. Of course. the value of your portfolio (excluding the short position in the option) is and you need to have .3) If the stock goes down. This option confers the right to buy the stock at time for dollars. and we can and do sometimes write rather than . and then borrow if necessary at interest rate to complete the purchase. You hedge your short shares of stock. the stock price rises at least as fast as the debt used to buy it. she should take to be the rate of interest for her activity. where is still to be determined. For example. since even in the worst case. and no one would invest in the stock. you dollars invested invest the residual money at interest rate . Suppose at time zero you sell the call for dollars. 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 ' $ —d’s! u ' $ —}’y! u  6 ¨ ' $ ! C … { z y x v XBA s |—d’yxb)uTw' DG¢ D C D ¢€¨ D D G¢ u # ‚llDEsDGWƒD u suT–1Ed5b4lGtd5s! u ' C ¢ ¨ $ ' # n ©$ n ' c$ ! C D ¢  ' c$ D ‚¢ Du ' c$ d5s! u A l' D C D W¨ D u suHt1†E‡$ ! C D ¢ ¢ $'# n ©$ n ' c ‚lsDElDG…–D u suHt1†„‘0b4lGt@0s! u ' C ¢ ¨ $ ' # n ©$ n ' e$ ! C D ¢  ' e$ ' e @0$ ! u D ¢ Du A ' D C D ¢ ¨ $ ' # n ©$ n ' e$ ! C D llEsGWƒD u suT–1E@&b4sG¢ # (1. the stock does not really go “down” if we get a tail. an agent is either borrowing or lending (not both) and knows in advance which she will be doing.2) ‚ F  # n ©  fHpH–o§ ¨ ' $ ! C$  ' $ s 4—d’yx5gt—d’s! u Du ¨ ! s –4C v ' s @EY5$ ¨ DCF # '! $ e–}’b! u Du D ¢ © © ¨ ! s –4C Du # 6 ks t # © r vW§ . (This is not as ridiculous as it first seems. let us consider a European call option with strike price and expiration time . you will have in the money market. We denote by the value (payoff) of this option at expiration. you don’t yet know which value will take. actually depends only on .) We assume that The model would not make sense if we did not have this condition. Thus. the value of your portfolio is and you need to have .

7) becomes Because we have taken .8) is not zero. both and are in the interval them as probabilities of and . this leads to the formula This is the arbitrage price for the European call option with payoff formula. They appeared when we solved the two equations (1. she means the derivative (in the sense of calculus) just described. respectively. Note.4) from (1. We want to determine the arbitrage price for this option at time zero. where and depend on and . At expiration. the first and second coin tosses. In fact.3) and (1. We now consider a European call which pays off dollars at time .3) or (1. and we solve them below Subtracting (1.7) as (1..6) (1.e. They are the risk-neutral probabilites.3). we obtain so that This is a discrete-time version of the famous “delta-hedging” formula for derivative securities.14 These are two equations in two unknowns.6) into either (1. Depending on how uncertainty enters the model. To simplify this so that (1. After some simplification.i. the denominator in (1. and have nothing to do with the actual probabilities of getting or on the coin tosses. both and are defined. the payoff of this option is .4). we define at time . at this point. To complete the solution of (1.7) (1.6) is a consequence of this definition.9).3) and (1. She then buys shares D G¢ D ‚¢ e c © '©A6 ss)$ ‹ Š F  gW§ ‹ Š ‰ ‰ ‚' e$ (ˆ‘0y! u ‹ Ed5s! u Š r‰ I© t© ‰ €D u n ' c$ Œ # n  ‰ ‰ §  ¨ A Š € uTWeoF F x  ‹ A WqHF –© x  Š ¨ © ' # § –© ¨ H¨ n $ § ‘# n ¨ ‰ ‰ ‰ © ! u § ¨ § ¨ # n ‚—ˆ'@0b! ' # q© oF ¨ n ' c $ § W# F n T–©  ‡ e$ u uHt1˜F Ed5y! u W‘H–© † © €D u n $ ¨ q ’ 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. however.2). Suppose an agent sells the option at time zero for dollars. 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. where is still to be determined.4) and solve for .5) (1.9) . and because they sum to .8) (1. we substitute (1. that my definition of is the number of shares of stock one holds at time zero. 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. This formula is so pervasive the when a practitioner says “delta”. and (1. not the definition of itself. . they are nothing more than a convenient tool for writing (1.4). we can regard Because of (1.

where is allowed to depend on because the agent knows what value has taken. investing a portfolio (excluding the short position in the option) valued at After the first coin toss. in the money market.CHAPTER 1.10) ’@€0uq u ' e e$ ‘d@&7q u ' c e$ ‘‘€7q u ' e c$  ' c c ‘dd5$ q u qu ! ¢ ! –’ (1. and depend on and . She invests the remainder of her wealth. she wants to have Although we do not indicate it in the notation.11). Considering all four possible outcomes. the outcome of © q x’ ! C! ¢ ¨ 4"`T–! ! ’ Ž ! ¢ ! Ž ‚'D CD ¢ ¨ lsEsGWƒD $ ' # n © $ n ' e $ ! C D x ' e $ u suTt1E@0bxiG¢ @0s! Ž A ' D C D ¢ ¨ $ ' # n ©$ n ' c$ ! C D yiEs‚W—D u suT–e„d5bxiG¢ d5y! Ž x ' c $ !–’ qu q PC ‚'! C! ¢ ¨ $'# n ©$ n q C! ¢  l"4B`Wt! Ž ibTh1tPB`ƒq u # n  ' e ‚(ˆ'@e•0$ q u ‹ Ed@0$ q u Š Œ H© –© 4@0$ ! Ž  e n' c e ‰ ‰ e $ q C ¨ ' c e$ q A '@€0ee 7PW¨4'dc@e0$7fC  ' e $ ! ' e $ ‘€0uq u 4d@07q u t@0s`¢ Du ! Ž ! 4C ‚ ' D C D ¢ ¨ $ ' # n © $ n ! C D ¢ x llEsGWƒD u suT–1rxsG@! Ž ' e$ ‘0b! Ž ' c$ ' e$ ! ' c$ ! D d5b! Ž @&b`¢ d5y“¢ G¢ D u D CD ¢ ¨ ElG…€D u (1. in the six unknowns . . . we can solve for ‚ '' e$ ! C' e$ ! ¢ ¨ ' e$ $ ' # n ©$ n ' e e$ q C' e$ ! l"@&i4"@&b`Wt@0l! Ž suHt1†E@€0uP"@&y`¢ A '' e$ ! C' e$ ! ¢ ¨ ' e$ $ ' # n ©$ n ' c e$ q C' e$ ! lw@0s4"@0s`…r‘0y! Ž suT–1Ed@0uP"@&y`¢ A'' c$ ! C' c$ ! ¢ ¨ ' c$ $ ' # n ©$ n ' e c$ q C' c$ ! lw‡y4w‡y`…4‡i! Ž suT–e„‘€bPw‡y`¢ A'' c C' c ¢ ¨ ' c $'# n ©$ n ' c c C' c l"d5$ ! "d5$ ! W4d5$ ! Ž suHt1Yd‡$ q w‡$ ! ¢ ' e$ @&y!`¢ ‚ ' e$ C' $ ! ¨ e$ l"'@0s!4"‘e0b`¢Wt'@&i! ' # © n ' e e $ q C ' e $ !  ' e e$ Ž $suTn–e$P@€0iP"@0b“¢ ’@€&bq u A ' e ! C e$ ¢ ' e $ ' n $ y'"@0$s4"'@&y!`…¨4@0$i! Ž su#Ht©1nE'd@0iP"@0b“¢ ”‡‘0bq u c e$ q C' e$ !  ' c e$ ' e @&$ ! ¢ d5$ ! ¢ D ¢ ' c . Suppose she decides to now hold shares of stock. At time .11) (1. the first coin toss. there are really two equations implicit in (1. the two represented by (1.10): . and . Therefore.12) (1. Introduction to Probability Theory 15 dollars in the money market to finance this. and .13) . we obtain the “delta-hedging for- ! –’ and therefore depend on Although we do not indicate it in the notation. To solve these equations. the agent has dollars and can readjust her hedge. . and she wants it to be . Thus. we begin with the last two Subtracting one of these from the other and solving for mula” and substituting this into either equation. the outcomes of the first two coin tosses. and thereby determine the arbitrage price at time zero of the option and the hedging portfolio .11) as four equations: We now have six equations. . we can write (1.10) and the four represented by (1. In the next period. the agent has of stock. her wealth will be given by the right-hand side of the following equation.

16 Equation (1. An example to keep in mind is (2.15)  ' c td5$ ! Ž (1.1) ™ of all possible outcomes of three coin tosses. The solution of these equations for and is the same as the solution of (1. Let be the set of all subsets of .13).17) (1.6) and (1.4).9).11) lead in a similar way to the formulas and . after the first tosses are known. Some sets in are . we plug the values and into the two equations implicit in (1. .3) and (1.16) This is again analgous to formula (1. postponed by one step.10). This formula is analgous to formula (1. denotes the value at The pattern emerging here persists. regardless of the number of periods. If time of a derivative security. but postponed by one step. We have just shown that (1.2 Finite Probability Spaces Let be a set with finitely many elements. and we denote it by . Finally. when the first . The first two equations implicit in (1. is given by coin tosses result in the 1. and itself. where should hold and the value at time outcomes of the derivative security. and this depends on the first coin tosses . the portfolio to hedge a short position shares of stock. where is the value of the option at time if . How many sets are there in ? ' e @0$ ! u (1.9).14).9). then at time . gives the value the hedging portfolio should have at time if the stock goes down between times and . .18) © ™  e ! —” ’A‚‚‚A! $ ˆ'‘"AsElt1"issl–d’X” u ‹ SdBsEyx1wssiihd’Q” u Š 4T© –© tB„yxe"sisl–d’bEl” u n ' cA! —” ’A‚‚‚A! $ Œ # n  '! —” ’A‚‚‚A! $! — ‰ ‰ ! —” ’A‚‚‚A! Eyx11isssh’ © ¨ ˜“ © ¨ ˜o“ eA! —” ’A‚‚‚A! $ ” C ¨ ' cA! —” ’A‚‚‚A! $ ” A '@wsElxe1ssis–d’XPq¨f'‡c˜As!E—l”x’eA"‚s‚i‚sAs!–}’$QPC  ! — ’ A ‚ ‚ ‚ !— ' eA! —” ’A‚‚‚A! $ @"sEyx1"isii–}’Q” u f‡˜sElxe"siss–}’7” u t' Ey” 1wssiA ! d’$ Ey” ¢ ”x’1"‚is‚iAs–’ A ‚ ! ”u  e e e  e e cA c e cA e c cA c c c „ ™ ›€€s… ›—w"w@j˜€w‡˜€ddb… š „  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@€1€w@1wd€1@€‰Bwˆw"€wdBwd‡u† „ ‚(ˆ'‘e€bq u ‹ E‡d5uq u Š fH© –© ƒd5s! u  c$ n ' c c $ Œ # n x ' c $ ‰ ‰ c—!–’ © ' c$ ‡y! u '@w5uPqfd‡7PC  ' c $ ! e c$ q C ¨ ' c c$ q ' e c$ ¨ ' c c$ ‘€bq u fd‡7q u 4‡b`¢ e– ! ’ @0$ ! Ž ' e ‚' e e$ ((@€0uq u ‹ Sd@07q u Š 4T© –© n ' c e$ Œ # n “ © ‰ ! —” ’A‚‚‚A! Elt1"isssh’ ‰ x€‘0y! u  ' e$ ' e$ ‘0y! u ' „y” e"sisA ! d’$ Ey” ¢ ! — ’A‚‚‚ !— © ¨ `“ D D G¢ ' e$  ' e$ ‘0y! u €u @0s! Ž ' c$ ‡y! u © ' c$  ' c$ d5y! u td5s! Ž 6 e  ! g—–’ “ © ¨ ˜`“ „ . We define this quantity to be the arbitrage value of the option at time if . defined by (1. and results again in (1.14) The hedger should choose her portfolio so that her wealth if agrees with defined by (1.

define for and for . ¯ Definition 1. ¥ A b™ q¥q 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 ! e c Example 1. Imagine that is the set of all possible outcomes of some random experiment. We think of as this probability.1 to determine for the remaining sets .2) (see Problem 1. Introduction to Probability Theory 17 (i) .CHAPTER 1. then ©A ˜i)6 Œ ™ !™ 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 ¥¦ ™ ¥¦ Definition 1. ¯ ¨ wGš (i) . In the above example. There is a certain probability. and then used Definition 1.1). A -algebra is a collection following three properties: c !™ ® ® Yƒ Eƒ ­ Eƒ Yƒ ­ p ® ® ® Eƒ ­   e e c A c e c A e c c A c c c … A ƒ ©  q p ­ © Un p ­ q © In ™ © V—ƒw"w@j"‘€d˜€ddbp ¥¦ of subsets of „ ' œ$ ‘5 ¥¦ ¬bª«© ‚Q€l…p ¥¦ ’  ' œ$ –‘5 ¥¦ ¢ ¡ ™ „ „ ™ ¨ j†œ ™ ¨ `œ For . it is generally the case that we specify a probability measure on only some of the subsets of and then use property (ii) of Definition 1.2) with the „ œ ™ !  P˜” !   fB” ‚ ' ” œ$ ž  ¡ ” ž  lu‘5 ¥¦ ˜l‘œ ‰g ¥¦ ¢ Ÿ œ q™ ™ (ii) If is a sequence of disjoint sets in . that when that experiment is performed. For the individual elements of (2. we define on the first toss is .4(ii)) to determine for all the other sets in . which is another way of saying that the probability of As in the above example. the outcome will lie in the set . we specified the probability measure only for the sets containing a single element.1 A probability measure properties: is a function mapping into with the following ‚ss‚iAs‘Bs›œ ‚ q œA! ©  $ t' „  ¥¦ © ' œ$ ‘5 ¥¦ ¥¦ 6 „ . Probability measures have the following interpretation. between and .2 Suppose a coin has probability in (2.2 Let be a nonempty set.1(ii) in the form (2. Let be a subset of . For example.

Suppose the coin is tossed three times. Similarly. The -algebra is said to contain the “information of the first toss”. you have been told that the first toss was a . for every set in whether or not the outcome is in that set. For example. let us define and let us define on the first two tosses so that We interpret -algebras as a record of information. Moreover. then ! ™ ¯ ¨ ° wƒ@œ · œ ¡ ¡ A „ «t² ³ Aš ‚i‚ssAu™•BAiq@˜i›œ ‚ œ œA! ¯ ¨ €†œ  g™  q º@™ ¡  D ™  ! ´™  q µ@™  ™ ¶€™ (ii) If . and nothing more. you might be told that the outcome is not in but is in . and you are not told the outcome. which is usually called the “information up to time ”. but you are told.2: To simplify notation a bit. contains the “information of ‚ ¹ œ · œ ¹ œ · Q S°¹ BA S°¹ BA ƒ°· ˜A ƒ°· œ A S¹ œ€± ƒ· ˜A S¹ œ€± —· BA Y¹ œw± ƒ· BA Y¹ w± —· BA ¹ BA · œ ¹ ¹ œ · ¹ œ ¹ · œ · œ · œ œ A ¹ ¹ œA · ¹ œA ¹ · œA · · œ A š xY€˜S•Brƒ•B‘—•BA „ )b… A Q on the first two tosses AQ A Q on the first two tosses A Q „ q ‘™ š ! ™ A ¹ œA · œ A š…  ! Q4€˜@•BA „ )b†–™ so that A  e e eA c e e…A  e c eA c c e…A  e e cA c e c…A  e c cA c c c… A š Q€€1w@€lyQ‘€€w€w@s˜Qƒw"w@jb˜Xjd"w‡dbyA „ «t² A ³ •€1€‘€1€wˆw€w@s˜Q›—w"w@j˜€jd"w‡dbyA „ «t²  e e eA c e eA e c eA c c e…A  e e cA c e cA e c cA c c c… A š and all sets which can be built by taking unions of these The set of all subsets of on the first two tosses A Q on the first toss A X on the first toss A ³ ¯ e …   e e e A c e e A e c e A c c e … x ¹ s¸—€ƒ"w€€1@w@w€d@i†q€œ c …   e e c A c e c A e c c A c c c … x · b¸|›€j˜€€w"€w‡˜€ddb†€œ „ ž ” !  @œ P˜” ± e e…   e e eA c e e… ¹ €s†—›€€"w@€i†x  S¹ œ c e…   e c eA c c e… · ‘s†–w@"wd@s†x  S¹ œ e c…   e e cA c e c… ¹ wb†–€wBw@wb†x  ƒ· œ c c …   e c c A c c c … x · · db†ƒjdBwddb†¸ƒ•œ © ¹ œ ‚ „ ¯ (iii) If is a sequence of sets in . e . then its complement .18 Here are some important -algebras of subsets of the set in Example 1. you would be told that the outcome is not in and is in . is also in . In effect.

and thus technically a random variable. if you see the first two tosses. The preimage under of this interval is defined to be is: f  ' $D †–—}’¾EC ‚ ©  ' e e e$ q C  ' c e e$ q 7t@€•0bftd@€0uPC ASf˜4'@ewc@0$7qPCP'dc‡c‘07P4@€w57Prd@w5uPC e e$ q C  ' e e c$ q C  ' c e c$ q A ½ ©  ' e c c$ q C  ' c c c$ q )s4@wd5uP4‡dd5uPC ¡ ™ RC q PC ™ Definition 1. is in .1) and consider the binomial asset pricing Example 1. and is denoted by . Introduction to Probability Theory 19 contains “full the first two tosses. whether is in the set. In particular.3 Let be given by (2. that is to tell you. where . A filtration is a sequence of -algebras such that each -algebra in the sequence contains all the sets contained by the previous -algebra. suppose the coin is tossed three times and you do not know the outcome . and are all random variables. and you know nothing more. albeit a degenerate one. for each set in not in . Then you know that in the first two tosses. for example. since for all . and we can look at the preimage under the random variable of sets in . This collection of sets is a -algebra. A random variable maps into . be the -algebra of all subsets of „ ¡ » ™A‚‚‚Aq ™A! ™AD €w"sisi@"B›"7€™ Definition 1. The so-called “trivial” -algebra contains no information. and even more. and is not in . the random variable of Example 1. A .CHAPTER 1. Nonetheless.” The -algebra information” about the outcome of all three tosses. More specifically. Note that defined earlier contains all the sets which are in . you can distinguish from .1. The information content of this -algebra is exactly the information learned by observing . and . ’ ¡ A ¹ ¹ œA · ¹ œ ± ¹ · œA · · œ A rY€˜S•j—•B‘ƒ€˜A „ )š ¥¼ „ ¹ œ  Àp à qC à ‚ Y°¹ @X7H‘PHGf ›„ Hl†—Âb7BSf Œ ƒ—d’uPC „ y… Á ¨ ’…   À pA ¨ ' $ q Á ¨ ’ we can get as preimages of sets in q PC 'q C$ BP–¡ qC ’ 'q C$ "P5–¡ ·¹ S•œ q PC q PC ¹ Y¹ œ ¹ · ƒ€œ f ' $q —d’bfC ’ ÀpA i7˜Sf Œ Let us consider the interval . . Knowing whether the outcome of the three tosses is in (it is not) and whether it is in (it is) tells you nothing about Example 1.3 Let be a nonempty finite set. The “random variable” is really not random.” which is the “information up to time . Then . For example. for example. called the -algebra generated by the random variable . We have The complete list of subsets of and sets which can be built by taking unions of these.1. You might be told. This means that the information in the first two tosses is greater than the information in . D ™ ™  ™ @€™ ¡ ¡ š ¡ p ¥¼ DEC „ q ! D PC 4C EC ¥¼ ’ „ ’ ¥¼ · œ ¹ S¹ j± ƒ· œ 'q C$ "P5h¡ q PC ¨ ¿’ ½ ©  D F  ' e c„ c$ q sv‘EC †t@wd57PC  pq  f  D ‘†F ”WEC !q ˜§ „ „ „ „ ¡ ¡ q @™ · ƒ· œ ¥¼ ¡ „ . but someone is willing . it is a function mapping into . but you cannot make this distinction from knowing the value of alone. Consider. there was a head and a tail. This information is the same you would have gotten by being told that the value of is . .4 Let be a nonempty finite set and let random variable is a function mapping into .

as in the case of above.) Important Note. defined by if if if if By the distribution of a random variable . the induced measure of a set tells us the probability that takes a value in . some sets in and their induced measures are: at the number . regardless of whether the probability for is or .20 Definition 1. For example. ©  ' e e e$ q C  ' c e e$ q ‘G@€€&bPkd@€&7PC ¼ Hœ ¥ Ä ½ s© In fact. finite set. Let be a random variable on . Definition 1. (Later we will consider random variables which have densities. A random variable is a mapping from to .-algebra of . or tell what the cumulative distribution function is. In order to work through the concept of a risk-neutral measure. a mass of size (2. we mean any of the several ways of characterizing . The -algebra generated by is defined to be the collection of all sets of the form .3) Ž œ ¥¦ ¼ ¥†Ägœ „ „ ¡ ¥¼ ¯ q fC Ž ÏÏÑ ‚ Í Ã ½ fHGs© 7© Ï A A½©  Í Ã )sjIf A Ê ÏÏÎ  ' Í Ã q C $ ¥ x  ' Í Af  Í Ã STjI© A ÊÒ ÏtYHP¦ €S$ Ì A©  7wÍ )Ë 6 Ð A È lÇ qPC 'Y5$ Ì Í ¥ ©  7™ ¤ È lÇ ËÊ q q ¥ qC !Ê  q ¦!™ ¤ ‚ ® ƒ ­Vt'r¹Y¹€$¦ ¥˜~uƒPup¦ SiB)6 Œ  œ ©  q C… ¥  ƒA q p A7©–' „ ¦ 4TB)6 Œ  $ ¥  ' ÉA A7©–' „ ¦ t¼ $  $ ¥  ' ¥ A 6  ' š$ ¥  ' š )t¦5¦ t7$ Ž  œ ¨ ' wt—d’$ Ž „ y… Á ¨ ’ „ ¥¼ Ž ¯ ¡ ‚ œ Q›€¨ Ž p¦ x  @$ … ¥ ' œ ¡ ' Ž –¡ $ œ ' h¡ $ ™ Ž !q ™ !™ œ ¥¼ œd¨t—d’$ Á ¨ ’ ' Ž wA ˜l… ¡ ' ™„ „ $  œ ›€¨ Ž … c Ž Æ hÅ „ œ ' ™ wA „ $ ÈlÇ Å ÈlÇ Å ÈlÇ Å ÈÇ l—Å œ ¯ „ „ f Ž q PC Ž Ž Æ –Å ËÊ ™ . If is discrete. in which case the induced measure of a set is the integral of the density over the set . where is a subset of . and a mass of size at the number . We say that is -measurable if every set in is also in .5 Let be a nonemtpy finite set and let be the -algebra of all subsets of . In the case of above with the probability measure of Example 1. A common way to record this information is to give the cumulative distribution function of .6 Let be a nonempty.2. Let be a sub. and let be a random variable on . It has an existence quite apart from discussion of probabilities. let be . Given any set . the induced measure of places a mass of size at the number . let be the -algebra of all subsets of . Note: We normally write simply rather than . we can either tell where the masses are and how large they are. in the discussion above. we a probabilty measure on define the induced measure of to be In other words. . we set up the definitions to make a clear distinction between random variables and their distributions. nothing more.

then the random variable has not. can take only finitely many values. If we set the probability assigns mass to the number . The payoffs of the call and the put are different random variables having the same distribution. a random variable can have more than one distribution (a “market” or “objective” distribution. The probability the payoff is is . (2. and let be a random variable on ..7 Let be a nonempty. Like the payoff of the call. a way of assigning probabilities The distribution of a random variable is a measure to sets in . The expected value of defined to be v BP…5$ 'q C ¨ ƒ is a into 6 „ „ „ !q „ p p ™Ë q PC ‚ ©  ' e e e$ q C  ' c e e$ q 7t@€•0bftd@€0uPC ASf˜4'@ewc@0$7qPCP'dc‡c‘07P4@€w57Prd@w5uPC e e$ q C  ' e e c$ q C  ' c e c$ q A ½ ©  ' e c c$ q C  ' c c c$ q )s4@wd5uP4‡dd5uPC e ¥¦ » ÍA‚‚‚A! )˜"sislEÍ ½ s© e c c  wd•’ „ ‚”Í QuYb… YÍ ” Æ hÅ  ”  ” Ž p ¥¦ ” Í Í … × Õª Y  7)©  ’… €lp ¥¦ ” YÍ Ø˜Ö t¢ Æ Ö × Õª Y  ¾«© €’yp¦ "'—d’$ Ž … ¥ Ø Ö ˜Ö t¢ Æ  ’… ¥' €lp¦ "—d’$ Ž ‚  ’… ¥' Q€lp¦ "—}’$ Ž ¡ ¥¼ Ë e e e  €™ •g’ p Ë p c c c ! d‡£’ p p ½s© !Ê Æ –Å c ™ CHAPTER 1. the probability of and the probability of is .1.4) as Ž ¥¦ Definition 1. finite set. It depends on the random variable and the probability measure we use in . let be the -algebra of all subsets of . It is still defined by Thus. Suppose in the binomial model of Example 1.on . Introduction to Probability Theory 21 Ôbª)© x  Ž ¥Ó ¢ Ž !fB”    »¢ !fB”    »¢ !fB”    »¢ !fB”    »¢ Ôbª)© x  Ž ¥Ó ¢ Y)́~» Ž l"issXlEt! Ž … » …A‚‚‚A ! Í  Ž È lÇ Å !Ë 6 Ë c e e  @•g’ ! p ƒ f l© ' ™ wA „ $ !™ c È lÇ Å „ v XlH¨ q 5$ 'f© C !q Ž ¥¼ p c be is . which we label . The payoff of the call at time is the random . Consider a European call with strike price expiring at time . two different random variables can have the same distribution. The payoff of the put at time is . which takes the value if or . and takes the value in variable every other case.e. or . The distribution of has changed. and the probability it is zero is . let a probabilty measure on . the payoff of the which takes the value if put is with probability and with probability .4) is defined to be a sum over the sample space . i. and then rewrite (2. and a “risk-neutral” distribution). We can partition the subsets . If we set the probability of to be . Since finite set. In a similar vein. then assigns mass to the number . Consider also a European put with strike price expiring at time . but of to be .4) „ Notice that the expected value in (2.

‚'” Í lsY5$ Var Æ hÅ One again. subsets of . although the expected value is defined as a sum over the sample space . we can rewrite (2. The definition of is with the distribution given by „ ¡ ¥¼ ™ Ž ¥¼ qC PG ¥Ó  qC ÙPG ¥Ó      „ ÍA‚‚‚ » B"isiA ! Í ¥¼ ¥¼ (2. then rather than over .5) . we consider (2. Indeed. finite set.5) as a sum over . This section concerns the Lebesgue integral on the space only.e. unlike the Riemann integral. but not all. This definition and the properties of measure determine the Lebesgue measure of many. is the collection of Borel sets defined below. let be a probabilty measure on . and for which Lebesgue measure is defined. i. it can be defined on abstract spaces. The collection of subsets of we consider. we consider the set of real numbers .8 Let be a nonempty. which is uncountably infinite. such as the space of infinite sequences of coin tosses or the space of paths of Brownian motion. we can also write it as a sum over .3 Lebesgue Measure and the Lebesgue Integral In this section. and let be a random variable on . let be the -algebra of all subsets of . the generalization to other spaces will be given later. We need this integral because. Var 1.22 Thus. Definition 1. We use Lebesgue measure to construct the Lebesgue integral.. The variance of is defined to be the expected value of . We define the Lebesgue measure of intervals in to be their length.3). a generalization of the Riemann integral. if takes the values ¥¦ Ž „ ‚ QÛÜ f Ú f Ú f Ú ½ fÛ ~n Ûf Tn ©Û ›s© © uQ… ƒ|«s… …–XsQ… ›s© Ú © n f Ú f n ½© Ú ½ u•Pbp¦ s•˜——˜ƒPbp¦ sƒÅ —XiƒPbyp¦ s›s© ©  q C … ¥ Ú © n yÇ )f Å  q C … ¥ Ú lÇ …n  ½ ©  q C È … —¥Å Ú ½ Ȑ Èf Ç 'r¹Y€¦ l€ES•jƒ€5¦ sTP@ƒ•5¦ s›s© ¹ œ$ ¥ Ú © n ' · ¹ œ ± ¹ · œ$ ¥ Ú f n ' · · œ$ ¥ Ú ½  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 Ir£‡‘sp¦ "dd@0$ q n e c e… ¥' e c e C n  c c e… ¥' c c e C ›e€wbp¦ "@•€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¦ "@wd57PTr£d‡up¦ "dd‡7PC Ž !  P˜” !   fB” ¥ ¨ ”Í  ”Í … ¥ ¨ ”Í  q ' Ž Ó W•S5$ » ¢ uS Ž p ¥¦ q ' Ž Ó W•S5$ » ¢ 4' Ž $ ¥¼ ¥¼ „ Ôª u)© ‚Q€’lp ¥¦ q ' Ž Ó ¥…¨P'—d’$ Ž $ … x €' Ž $ ¢ q' Ž Ó ¥W¨ '`$ "A $ Ž™ „ q PC ¥¼ „ To make the above set of equations absolutely clear.

In addition. so ' ÉA ß$ ± ' ÞA É ¨ T˜u5WPXB˜€1$ ß Þ For real numbers and . Since is a -algebra. Since is By definition. where and are real numbers. denoted intervals in . as is This shows that every closed interval is Borel. For example. the union contains Þ ' ¥$ ¼ 5~Ý ' ¥$ ¼ 5Ý 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. Indeed. ¡ ß !  Pu» ‚'ÞA à ¨ Þ lQ˜4W5$ ž tXB˜€1$  'ÞA É ¨ Ÿ !   fu» ' à n ÞA Þ „T•BY5$ ž rT˜S$  ' ÉA Þ Ÿ Þ ' ¥$ ¼ 5Ý ' ¥$ ¼ 5Ý ' ¥$ ¼ ~Ý 'ßA Þ lBY5$ ' ¥$ ¼ 5Ý ¡ ¡ ¥¼ . the closed half-lines and are Borel.  » ˜"sisA q ˜A ! b㜠ÞA‚‚‚ Þ Þ…  This means that every set containing finitely many real numbers is Borel.3. every open interval a -algebra. if then Þ Every set which contains only one real number is Borel. For example. since they can be written as intersections of open half-lines and closed half-lines. the open half-line is a Borel set. The sets in are called Borel sets. if is a real number.CHAPTER 1. every union of open intervals is also in . then ' ¥$ ¼ 5Ý ‚ ° ¥ T˜u5WPXB˜€1b§%wBYÞ Œ ' ÉA ß$ ± ' ÞA É ¨$ ¤  ßA !   fb» ‚ ® © à €BA © à pQ­ ž @«b… n Þ ¨ Þ  Þ ⠂ ' ÉA Þ$ á ßA É ¨$  ßA Þ lT˜S5WQw˜˜€1gSwBY5$ !   fb» ÞA à ¨ ¾B4WpÞ Œ ž „¾B˜€e$  ÞA É ¨ Ÿ !   fb»  à n ÞA «I€BYÞ Œ ž tTBYÞ Œ  ' ÉA Ÿ ‚”Þ XuXb… !f˜”    Hœ »Ÿ ¡ ' ¥$ ¼ ~Ý is Borel.9 The Borel -algebra. is the smallest -algebra containing all open Definition 1. for every real number . The following ¡ . every complement of a Borel set is Borel. Introduction to Probability Theory 23 . Half-open and half-closed intervals are also Borel. is in . .

sets which are not Borel. We use it later when we discuss the sample space for an infinite sequence of coin tosses.” there are lots of points in this set. Thus. Thus. so at stage . We shall see eventually that the Cantor set has uncountably many points. We have just seen that any non-Borel set must have uncountably many points. the points are all in . Continue this process. This is a countably infinite set of points. In particular. has zero “length. has two pieces. remove the middle half. and in general. and each piece has ‚‚‚Aq ÞA! ޅ  1sisiQ˜i«b†˜œ In fact.e.e. The “length” of the next set removed is . the length of the is -th set removed is .. ç !  P˜” ” ž x å â å ” “ ”p ½ f s© ½ ½ s© A  å s© f ‚—‡˜iA s© © ½ æ s© † Ÿ ‡ s© A ƒ † Ÿ ‡ © A ƒ † Ÿ ‡ © )6 † ƒq ƒ å ½ ½ s s ‚ ® s©© A s©© ­ ® sƒ½ © A s©½ © ­ ‘œ x q æ ƒ Ÿ q @œ ½s s ½ ‚i‚ssA u©f ½ A æs©© A s©© ‚ ƒ ‚‚‚ ssiA q Ú !Ò  ˜!q ™ f q ! !   fB” A© ¦ ” © p ž ¢ A sƒ½ © A i©½ © 7iA ƒf A © f )6 A© A A! å fA  © )6 † –! å The remaining set ‡ ©iA f ƒ† ± ‡ ® ‚ Pƒf A © f ­ x ! äœ å ©A ˜i)6 Œ Consider the unit interval . There are. i. the first set removed. remove the open interval has pieces. . every set containing countably infinitely many numbers is Borel..4 (The Cantor set. Example 1. The “length” of . is . if . the second set removed. From each of these pieces.” Despite the fact that the Cantor set has no “length. the set length . none of the endpoints of the pieces of the sets is ever removed. then !f˜”   ‚Xu”XÞb…  Hœ »Ÿ ! ›œ ” «— p !Ë  !Ò n !Ò å Ö !Ë “ .) This example gives a hint of how complicated a Borel set can be. i. and remove the middle half. the set of points not removed. The Cantor set is defined to be the set of points not removed at any stage of this nonterminating process. as is its complement. Note that the length of .24 This means that the set of rational numbers is Borel. however. the set of irrational numbers. remove the open set The remaining set has four pieces. the total length removed is and so the Cantor set.

because of the “length” computation given at the end of that example. In fact. and property (v) in Problem 1. Introduction to Probability Theory 25 Lebesgue measure is defined to be the measure on which assigns the measure of each interval to be its length.4 must be zero. Following Williams’s book. then ¥¼ ¡  ÉA £˜)6 Œ ' ¥$ ¼ 5~Ý Ý Definition 1. For example.4. but . then '' ¥$ ÝA ¥ "¼ 5~wG¼ 5$ . we obtain ‚ p à  ® © à ‘BA © à GÞ ­ ET«buETG6 n Þ ¨ Dè à Þ…Dè à ® à n Þ à ¨ Ä Þ © •BA © pÞ ­ ï«b… à ž š  ” !  ‘œ P˜” á Then . ). except that the total measure of the space is not necessarily (in fact.4 needs to be modified to say: To see that the additional requirment is needed in (v). so . is a sequence of disjoint sets in . The Lebesgue measure of a set containing only one point must be zero. and that determines the Lebesgue measure of all other Borel sets. the Lebesgue measure of the Cantor set in Example 1. A measure on is a è . A measure has all the properties of a probability measure given in Problem 1.10 Let function mapping be the -algebra of Borel subsets of into with the following properties: .CHAPTER 1. we must have ‚6  Þ…D «@)uu„è É ð ‘jà Letting . É  '! œ$ HƒB›5–è ž ž ɁtQ»€Q„è boíyìëê  ' œ$ D î» 6  ' ” !   á$ D tu‘œ P˜” 5QEè ‚ ‚ A' É 1‚is‚syTBA)ƒ Œ –€˜lTB)p Œ t@˜lTB¦© Œ —›œ ™ œA' ÉA  q œA' ÉA  ! É  '! œ$ HƒB›5–è ž !   fB” ‚'» œ l)€$|è íyìëb»  ¡ ”‘œ ž  è êî º â ' ¥$ ¼ 5~Ý ÚÚÚ é q œ é ! si‘@H‘œ ‚‚‚Aq œA! sisi@˜sœ (v) If is a sequence of sets in with and Dè !P ˜” !   fB” ‚ œ$ l'i”@5–è ž ¡ ‘œ ž  è ” ¢ Ÿ ' ¥$ ¼ 5~Ý (ii) If '' ¥$ ÝA ¥ "¼ 5~wG¼ 5$ ' œ$ è ¨ ©  œ$ ‘5–d€t' ° 5–è É  ' ¥$ D t¼ 5QEè © ‚ss‚isq‘Bs›œ ‚ A œA! 6  ' š$ t¦5–è (i) .4(iii). one no longer has the equation in Problem 1. we denote Lebesgue measure by . consider We specify that the Lebesgue measure of each interval is its length. since for every positive integer .

and we shall pretend such functions don’t exist. We may not compute the Lebesgue measure of an uncountable set by adding up the Lebesgue measure of its individual members. possibly taking the value is simple and ö ” and each is a real number. then The Lebesgue measure of a set containing uncountably many points can be either zero. In order to think about Lebesgue integrals. We for every 6 ‚' œ$ D è D è § l@Q„gx  E¦ˆò õ Xó ô ò ¾©r'Y5$rò Á ¼ €‰b†œ Í ¥ ¨ Í … x ¥¼ ¥¼ ò Definition 1. we want the -algebra generated by to be contained in . a function of the form to be at some points. It is difficult to conceive of a function which is not Borel-measurable. !f ˜” !   f˜” ‚)66 ž~uXbuEè ž t‘5QEè   ”Þ…D  ' œ$ D ¢ ¢ ò  œ ¨ ' Í$ ¥ ¨ Í …Sxñ Á ¼ q£b… ‚‚‚ Þ Þ…  1sisA q ˜A ! b†˜œ ¥¼ ¥¼ © ñ ¡ .11 Let be a function from to . Hencefore. Definition 3. positive and finite. the set indicated by . Indeed. “function mapping to ” will mean “Borel-measurable function mapping to ” and “subset of ” will mean “Borel subset of ”. we must first consider the functions to be integrated.e. We call from to is a function which takes only the values ñ ¥¼ ¥¼ ' ¥$ ¼ ~Ý ' ¥$ Ý ¨ ¼ 5Wvœ ¥¼ ¥¼ ¥¼ ¥¼ ' ¥$ ¼ 5Ý ñ ñ Definition 1. i. or infinite.. if . because there is no way to add up uncountably many numbers.26 The Lebesgue measure of a set containing countably many points must also be zero. We define the Lebesgue integral of to be Let be a nonnegative function defined on define the Lebesgue integral of to be ‚ü¼ ¥w¨€Í ' Í$ ñ à ' Í$ Y5tHƒSrö ö Dè§ û ú ù ø x D è § Á „¦‘ö õ 7ó p)B˜E7ñ õ 7ó ô ô ñ É ¥¼ ÷%!f˜”   ÷!  )P˜” ‚ œ l'i”@$QD„èy”  D觔 @E7iQò ” x D è § E7‘ö õô ó » ¢ õó »¢ ÷ . We define the Lebesgue integral of AX‘“…Í ” œ h¨ A” œ ¨ X‘“wÍ A)6 A ²  ' Í$ ” 7© gtY5QQò ” Qò where each is of the form if if A ' Í$ ” ò lY5QQi” ÷%!fB”    ' Í$ tS54ö »¢ ¥¼ ¥¼ ö A simple function from to is a linear combination of indicators. We say that is Borel-measurable if the set is in whenever . The integral was invented to get around this problem.4 is purely technical and has nothing to do with keeping track of information. In the language of Section 2.12 An indicator function and .

and so the Lebesgue integral of over To compute the Riemann integral . has Lebesgue measure zero. The Lebesgue integral just defined is related to the Riemann integral in one very important way: if the Riemann integral is defined. since (or equivalently. Being a countable is .5 Let be the set of rational numbers in . Introduction to Probability Theory 27 is integrable. and consider set. where . we say that ”‹ ©A ˜i)6 Œ Í §' Í$ QwS5tñ ! D þ ó Í §' Í$ ñ X"S5t€£ þ œ ¢  ñ ©  » ’˜)Í ”# ñ  É €¨ are finite at . then the Lebesgue integral agrees with the Riemann integral. Let be a subset of . In each subinterval there is a rational point . If both ). let be a function defined on . we say that and is integrable. possibly taking the value other points. Let be a function defined on . we choose partition points and divide the interval into subintervals .CHAPTER 1. as we show in the following examples. We define is the indicator function of . possibly taking the value at some points and the value at other points. We approximate the Riemann integral from above by the upper sum and we also approximate it from below by the lower sum © ‘s” ‹ xñ  ' $  » ÍA ! — » A‚ ‚ ‚ Aq ÍA! A! ÍA D i)BsE¦)Í Œ (isi(1YBBRÍ Œ 1ÂRBu)Í Œ  ÚÚÚ  ! Í  DÍ  tsiVEUw)V€6 Dè§ ¦ ¤ „¦‘ñ ©¨¢ §£ ¥þ ©A ˜s)6 Œ  x ¥ ‡ñ   !   fB” !fB”   A7©t' !Ey” Íq¨ ” 5$Xڗ© Í h' !El” Íq¨ ” ́s' ” ‹ tñ $ $ — — »¢ »¢ 6h' ” 5tñ #$ ” ÍA! —” iSBsEySÍ Œ ‚ 6  '! — ” Í ¨ ” Í$ Ú )tBEySq•S5X@6 ñ Ap€…Í œ h¨ A œ ¨ p€wÍ ©A ˜s«6 Œ ‚6  Dè§ )ˆE¦ñ ¨ ! ¦ D ¤ !fB”   !   fB”  '! — ” Í ¨ ” Í$ ' ” #$ hBElYW€Sssi5tñ »¢ »¢ A)6 A ² x ' Í 7© g€S$ ¬  ¥ where x  D ¦‘ñ ¬ A D ¦ñ ¬ ¥ è§ è§ ó   ¡õ ô ó ¥¼ œ ɀ¨ É ¥¼ ñ ñ vñ É  Dè H‘„¦§ ñ þ — Tn ýÉ ñ DE7§ — ñ þ è Dè E7§ v ñ þ É w¨ ÿ ÿ ÿ €õ ô ÿ õô ADè uE7§ õ ô ¨ ¨ Dè Hƒ„¦§ v ñ x D è § ïE¦@ñ —ñ õô ó õô ó õô ó ñ AQX6By'S5$tñ€QYu€Y5$ — ƒQX˜lSxu)u€Y5$ v ñ A Í ¨ … { z y x ' Í ñ A  6 A ' Í $ ñ … { z y x ' Í ñ ¥¼ to be . at some points and the value if if É ñ It is possible that this integral is infinite. If it is finite. and there is also an irrational point . where . Example 1. Finally. The first is that the Lebesgue integral is defined for more functions. The Lebesgue integral has two important advantages over the Riemann integral. We define the positive and negative parts of to be respectively. and we define the Lebesgue integral of provided the right-hand side is not of the form .

the upper sum is always and the lower sum is No matter how fine we take the partition of always . Every simple ç © A )6 x  ' Í $ A ˜É ² •S5tñ A )6 ² x  ' Í $ A g‘S54ö ©A ˜s)6 Œ ñ 6 É õô ó 6 6 É . and again the Riemann integral does not exist. if and ' Í$ ò à ' Í$ Sf…ƒY5tñ and whenever for all . In particular. one of these will contain Riemann integral the point . On the other hand. The Lebesgue integral has all linearity and comparison properties one would expect of an integral. for any two functions and and any real constant . we have are disjoint sets.6 Consider the function It follows that Now consider the Riemann integral . and when we compute the upper approximating sum for . When we partition into subintervals. then ç ‚6 « ü ¼ €€Í ¥ ¨ Í § ' Í$ ! QYS5tñ E! — þ ñ A D 7ñ è§ õô ó A7DEè7§€ò n Dè§ –E¦@ñ ô õ ¦ó ô ÷ õ 7ó ' Í$ ñ à ' Í$ S5tHƒY5rö is simple and ‚6  6…D  Dè§ )@XbuEè  €E7‘ö ‚ § sDE覧—Sò Ã~DE¦ñ è§ õ 7ó ô õ 7ó ô ¼ €€Í ¥ ¨ ÷ ÷  D 7ñ è§ õô ó  Dè§'ò n ñ µE¦SS£G5$ ©A© ˜s¦¨ Œ 6 ž Í § ' Í$ XSY5tñ ž — þ ò õô ó ö Á E7‘ö Dè§ ñ õ 7ó ô õô ó Í § ' Í$ ! QYS5tñ E! — þ û )B˜€E7ñ ú ùø  D è§ ' ÉA TB)6 Œ ¨  for some . this point will contribute times the length of the subinterval containing it. Since these two do not converge to a common value as the partition becomes finer. Thus the upper approximating sum is . the Riemann integral is not defined.28 . which for this function is the same as the . and thus has Lebesgue integral A)Í 6  A6  )`Í if if for every É This is not a simple function because simple function cannot take the value function which lies between and is of the form ‚)Í 6  A6  )Í if if . Example 1. ‚Dè§ 7„¦ñ ! –E¦ñ ¬ n Dè§ ó ó  Dè§ ! ˆE¦ñ " ¬ ó  œ Finally. the lower approximating sum is .

where ‚Dè§ » 7E¦—iñ A©  Dè§ » 7€E¦—iñ õ Qó ô ž î ’Ã D è § íyìëb» E7ñ ê õ 7ó ô ž î êëy b» õ 7ó ô Dè§ » E7ƒsñ þ ž î» bo êëy If is defined.CHAPTER 1.7 Consider a sequence of normal densities. the Lebesgue integral of this function was already seen in Example 1.7 and 1.8.8 Consider a sequence of normal densities. These converge pointwise to the function Example 1.1 (Fatou’s Lemma) Let verging pointwise to a function . Then be a sequence of nonnegative functions con- ç 6  Dè§ ýd„¦ñ þ õô ž ©”€DEè7§)»iñ þ boíyìëê î» à ‚)Í 6 õ ô  A )6 A6  )Í ˜É •S5tñ A ² x ' Í $ if if ç à 6 6  Dè§ @E¦@ñ þ 9 ë ê à Dè§ "87Tbíyî » ë TE7ñ ž õô ó ñ ©  D觻 ‘dE7)iñ þ ©  D觻 @E7Xsñ þ We have for every . then Fatou’s Lemma has the simpler conclusion ‚Dè§ » 7„¦—iñ õô ó ‚‚‚A pA©  àA » lss«B7–dBEiñ Theorem 3. Pointwise convergence just means that There are no such theorems for the Riemann integral. The function is not the Dirac delta.6 to be zero. so . à © Example 1. In each of these the situation is that there is a sequence of functions converging pointwise to a limiting function . each with mean and the -th having variance : These converge pointwise to the function We have again for every . This is the case in Examples 1. because the Riemann integral of the limiting function is too often not defined. Introduction to Probability Theory 29 There are three convergence theorems satisfied by the Lebesgue integral. but õô & %p x  ' Í $ ‚ 2 ) — à €S5$ » ñ 6 È È ž ©  D觻 €E7)iñ þ b`íy êë î» à ‚ ¥ ¨ G¼ €wÍ õ ô tS5tñ 6  ' Í$ for every . Before we state the theorems. so . each with variance mean : ‚ ¥ ¨ G¼ w€Í and the -th having ‚‚‚A pA  à lsi)B¦© £BA » ñ for every & $ '%p x  ' Í $ » ‚ 530)'— © # €Y5„iñ 421 ( È È ž î êë ' Í$ ñ  ' Í Y5ttS$ » ñ íy b» õô ñ õô õô ñ ñ à !» . but . we given two examples of pointwise convergence which arise in probability theory.

is and the larger one. converging pointwise to a function . ) such that Then and both sides will be finite. . Fatou’s 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. the smaller one. ™ ¨ jœ ž î êº § A7DE覧—iñ » íyìëb» €DEè7ñ õ ôQó õ 7ó ô ' Í$ ò ' Í$ » Sf…à S5„iñ for every for every consists of three objects: ñ ‚‚‚ p  A» isiA)BA7©ãà˜Siõñ ô É ž î êº D è AD § 7E覗»iñ íyìëb» €E7§ñ õ ôQó õ 7ó ô ‚G¼ ¥€¨jÍ Ú Ú sÚiÃ'S5Q™iHÃY5$7quñHƒSb7IG6 Í$ ñ ' Í Ã ' Í$ ! ñ à ñ .30 . a probability measure on . a -algebra of subsets of .e. .4 General Probability Spaces Definition 1. a function which assigns to each set a number . We could modify either Example 1. Now if is even.2 (Monotone Convergence Theorem) Let converging pointwise to a function .7 or 1. called the sample space. and . The has two cluster points.. Fatou’s Lemma guarantees that even the smaller cluster point will be greater than or equal to the integral of the limiting function. which represents the probability that the outcome of the random experiment lies in the set . . but if is odd. Then where both sides are allowed to be Theorem 3. By definition. Assume that there is a nonnegative integrable function (i. which contains all possible outcomes of some random experiment. if is odd. Assume that be a sequence of functions for every © à ž î ú ùø D„覧—7ò þ b» )Bíyìëê p » © õô p à ä D 7§ » ò þ © è ‚Dè§ » 7E¦—iñ õô ó ž î º D è § íyìëb» €E7ñ ê ' ¥A ™ ¦ ˜€"A „ $ ò ' ™ `"A „ $ õô ó ÿ „ ÿ ž DE¦—7ò þ ui"87Tíy êë è§ » î»9 ë !f bž»  è§ » þ uDE¦—7ò õ ô b… à » 7p  » ò ñ 6  D èõ ô § ïE¦ñ þ œ © A ¨ ' œ$ ˜s«6 Œ ‘5 ¥¦ ¡ ¥¦ ™ „ while but sequence à U D ¦§ » ò þ p è »ñ  » igj¾ò õô õô õô . ‚4à ¼ €wÍ ¥ ¨ É  Dè§ HE7€ò þ ‚‚‚A pA  àA » isi)B7© dBSiñ Theorem 3. The key assumption in Fatou’s Lemma is that all the functions take only nonnegative values.3 (Dominated Convergence Theorem) Let be a sequence of functions.8 by setting if is even. is . a nonempty set. 1..13 A probability space (i) (ii) (iii) . i.e. which may take either positive or negative values. .

and so give it a name: . Toss a coin times. Introduction to Probability Theory 31 has the following In particular. so that is the set of all sequences of and which have components. We repeat these and give some examples of infinite probability spaces as well.4 that a probability measure properties: . and so only finitely many terms appear in the sum on the right-hand side of (4. then with ™ à (c) (Finite additivity) If is a positive integer and ‚l' » $¦ ¥TBisE' ! 5¦ t' » jBis£± ! 5 ¥¦ œ nÚ Ú Ú n œ$ ¥  œ ±Ú Ú Ú œ$ » œA‚‚‚A! €˜"siss›œ !  P˜” !   fB” ‚ ' ” œ$ ž  ¡ ” ž  lu‘5 ¥¦ ˜l‘œ ‰g ¥¦ ¢ Ÿ are disjoint sets in ™ ‚‚‚ œ isiA q BA ! œ (b) (Countable additivity) If 6  ' š$ t75 ¥¦ (a) . (d) (Continuity from below. we define We can define this way because has only finitely many elements. then . Suppose the probability of on each toss is .1) . then ¥¦ Remark 1. is a sequence of disjoint sets in .) If is a sequence of sets in with ÚÚÚ Ä q œ Ä ! ss@‘H€›œ à ‚ © » ¡¹ T WPVAHA¢ Ga U‹ Ú © » · SQAA¢ Ga EŠ ï€lp ¥¦ T R P I H F D x  ’ … R I F D ' » ’A‚‚‚Aq ’A! $  Qr1wssist1s–d’£’ ¨ © Š —ãx  ‹ e ž   î êº ¡ ” !fB”  ‚ » œ$ l'S•5 ¥¦ íyìëb» ˜l‘œ ž g ¥¦ ⠙ ‚‚‚ œ isiA q BA ! œ ž î º ” !f B” ‚'» œ lS•5$ ¥¦ íyìëb»  ¡ ‘œ ž  ¥¦ ê Ÿ ‚‚‚Aq œA! isis‘Bs›œ ™ ‚ ' œ B $ ¥ n ' œ$ ¥  ' $ l‘€CA¦ IP@¦ ƒA ¥¦ is a sequence of sets in » „ ™ c ‚ ' œ$ ¥ r ' $ l‘5¦ H•A ¥¦ » „ Š  Ä @œ c „ ™ e  à œ (d) If and are sets in and .) If We have already seen some examples of finite probability spaces.1). Example 1. We will use this space quite a bit. Let be the collection of all subsets of . For each in . a number between zero and one.1 We recall from Homework Problem 1.9 Finite coin toss space. ç ¬bª«© x  ‘5 ¥¦ ‚Q€l…p ¥¦ ’ ' œ$ ¢ œ ' œ$ ‘5 ¥¦ ™ ¨ j†œ For each . then . we define ÚÚÚ isé q Hé ! œ œ (d) (Continuity from above. Then the probability of is .CHAPTER 1. then » „ (4.

so that is the set of all nonterminating sequences of and . However. and we need to exercise some care in the construction of the -algebra we will use here. we put every set in every -algebra integers. We also put in every other set which is required to make be a -algebra. and then we have In other words. the probability of a on the second toss is .  à on every toss !   fb» c u… ž   â c b… ™ c » €™ is in and hence in .10 Infinite coin toss space. for each positive integer . and it is clear how to define . We call this space . the set We next construct the probability measure on which corresponds to probability for and probability for . we define example. For c „ ¡ ž x  x  Y€œ ¹¹  x  S•œ ·¹  x  ƒ•œ ¹ ·  „ à · · —•œ q ‘™ ¡ ©A ˜i)6 Œ e œ . then the description of depends on only the first tosses. For example. . suppose .32 Example 1. If there is a positive integer such that . This is an uncountably infinite space. because it depends on all the components of the sequence and not just the first components. . is a -algebra. 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 w‡c The set of all sequences which begin with à  c  q ’A c  ! '‚‚‚ A ™ ’A q ’A! $  ’ £ƒx1wt–’ Á ˆssib41uxei–}’gql…  c  q ’A e  ! '‚‚‚ A ™ ’A q ’A! $  ’ W–xe@˜t–’ Á ˆssib41uxei–}’gql… e˜ƒqx1Awct!–’ Á ˆssib41uxei–}’gql… ’ '‚‚‚ A ™ ’A q ’A! $  ’  e  q ’A e  ! '‚‚‚ A ™ ’A q ’A! $  ’ g–xe@˜t–’ Á ˆssib41uxei–}’gql… „ š ¡ » €™ For each positive integer . Therefore. Let be given. We set and . where these sets were defined earlier. For example. to be the -algebra determined by the first tosses. the set containing the single sequence is not in any of the -algebras. we must also put into it the sets . Toss a coin repeatedly without stopping. Then is in . on the first  à on the first tosses tosses  on every toss ¡ à ™ à » ™ c…  ÚÚÚ c c c c c b†~ess†ddd‡u… ¡ c b… ¡ » ™ ™ ¡ à q @™ Because sets. where ranges over the positive In the -algebra . contains four basic sets. à ¨ Š c Š ‚ Š  Š ' ‹ n Š $g ˜‹ n q Š t@S•€Tƒ€5 ¥¦ h@ ¥¦  ' · ¹ œ ± · · œ$  ' œ$ Š  ' · ¹ œ$  ' · · œ$ q @™ Š˜‹ tS•5 ¥¦ Š –—•5 ¥¦ ' œ$ ‘5 ¥¦ · œ · œ  Y¹ G± ƒ· ˜œ q œ » ™ ¨ €jœ à ž ™W¨vœ '`"A „ $ ™ ¥¦ e ¨ Š @©  ‹ ™ is also in .

we define a sequence of random variables !  P˜” ‚ —d’” Xp „8  ' ' $ ” » ¢ 4—d’$ Ž ž „ 6 ² x ' $ ” © •—d’QE8 Ž c… bp ¥¦ If . otherwise. are those which contain uncountably many the only sets which can have positive probabilty in elements.1) corresponds to two sequences ž The numbers in (0. then every set in which contains and ) has probability zero. on every toss à e c ©  ©  Š G6  ‹ G6 c… bp ¥¦ ©  @ ž î  c… up ¥¦ íyë u» @ ê on every toss on the first tosses ž „ 6  @ ž îê »  ‚ » Š íyë bº@ ‚ Q c…  b† à on the first tosses on every toss  on the first three tosses  AYsÚsÚ Ú cb… c b… é  on the first toss on the first two tosses . . Indeed. A similar argument shows that if so that . Every dyadic rational in (0. For example.1. „ ¨H’ ‚ ƒf t1siw‡dd@w5$ Ž t1si€€€€•€d5$ Ž  'ÚÚÚ c c c c c e c  'ÚÚÚ e e e e e c c ž ¨ „ ’ ©A is)6 Œ X Ž “ Ö F q ©ºd(Úisځ‡dd5$ Ž ' Ú c c c c ” E8 Since each ™Ë 6  'ÚÚÚ e e e e t(isd€€€&$ Ž A e  ” @˜‘x’ A c € ” ’ if if ‚‚‚Aq 8A! issi„wsP8 In the infinite coin toss space. but sets can have positive measure. where and are integers.CHAPTER 1. For example. and we also define the random variable is either zero or one. We are in a case very similar to Lebesgue measure: every point has measure zero. and the other values of lie in between. we write them in terms of sets which are in for positive integers . For example. these numbers ™ ¨ …vœ à à é é  ™ q¨ œ » €™ !   fu» c b… ž â c u… c u… c… up ¥¦ ©  Š . and hence very set only one element (nonterminating sequence of which contains countably many elements also has probabiliy zero. and then use the properties of probability measures listed in Remark 1.1) which are not dyadic rationals correspond to a single have a unique binary expansion. is a dyadic . We define a “dyadic rational number” to be a number of the form . by . rational. Of course. Introduction to Probability Theory 33 for which there is no positive integer such that . then on every toss . and According to Remark 1. To determine the probability of these sets.1(d) (continuity from above). Such Let us now consider a set is the case for the set on every toss . takes values in the interval .

but the sequence must begin with or .2. Therefore. we can verify that for any positive integers we have In other words. 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. other than !q Æ hÅ X “ ‚ © f @£dc   c   e A © f @wc A © f @W‘e   c A © f @e•e A ©p @Wc   A ©p @e   and Š ©A ˜i)6 Œ p ‚ ” © p ›‡ ” Xp A I” vX  © ¨ † Æ –Å p A© ¦à ” Xp  H” vX G6 © ¨ à q  ‹  Š ' ™ "A „ $ ! ¥¦ … ¥  ‡© p¦ ~˜sA f ƒ† … ¥ p¦  ‡ f p ƒA ©† …p¦  ‡ p f ¥ ©A ©† …p¦ ï‡ f A ¥  © «6 † … ¥  ‡© p¦ ~˜sA p ©† …p¦ ¥ï‡ p A  © «6 † Æ Å Æ hÅ Æ hÅ Æ Å Æ hÅ Æ hÅ Æ hÅ Æ –Å ©A ˜s«6 Œ e wc „ c @e Æ Å on . the numbers between can be written as . Let be the subset of in which every even-numbered toss is the same as the odd-numbered toss immediately preceding it. but is not. the -measure of all intervals in whose endpoints are dyadic rationals is the same as the Lebesgue measure of these intervals. Similarly. We conclude this example with another look at the Cantor set of Example 3. then we have satisfying when we .34 Whenever we place a probability measure on . is the beginning of a sequence in . Ä 3 å å ' e™d ! A e!d ! $ ç abI T £ `„ Y e wc a Y bI T £ c„ ˜sA«6 Œ ©  3 å å 3 å 3 ew@€e c e c‘e€e€e ’ ' —d’$ Ž å å 3 ’ ' !q A ! Ë $ å '—d’$ Ž ‚Q abI T £ ¨ Yc„ q’ Á '—}’$ Ž …†x  3 å c c e e e e c d@€•€w‡c „ ¥¦ It is interesing to consider what construct the probability measure would look like if we take a value of on . we are removing exactly those points in which were removed in the Cantor set construction of Example 3.e. . none of these numbers is in . Consider now the set of real numbers The numbers between can be written as . The only way this can be is for to be Lebesgue measure.2. 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. i. by requiring consecutive coin tosses to be paired. one can convince onself that in fact . With a bit more work. but the sequence must begin with either or . In other words. For example. so none of these numbers is in . we have a corresponding induced measure in the construction of this example.

Since there are infinitely mean numbers in . we can speak of the probability that the number chosen lies in a particular set. Nonetheless. and setting up the probability measure so that The probability measure described in this example is . and for every Borel set .10 and 1. ç ç ©A ˜s«6 Œ Œ ©A ˜i)6 ½ s© ©A is)6 Œ f © ËÊ ½ s© © ‚ ©Û ~XiQp¦ €A Ûf @«sp¦ @A Ûf uQp ¥¦  ½©… ¥  f… ¥  ©… !™ c ‚ p — qÍ È yÇ Å f & $ f"p ©  ©A 㫘i)6 Œ D è '‘5$%DEè˜h'@$ ¥¦ œ œ ©A Ä ˜s«6 Œ †œ ' © A $ Ý  &is)6 Œ ~§™ ©A ˜s)6 Œ  „ !Ê x ' Í $ © # •S5€2 © ' ¥$ Ý ¨ ¼ ~dœ ½ s© ©A ˜i)6 Œ ¥¼ f  D gˆRC ' ¥$ ¼ 5Ý ' ¥$ Ý  ¼ 5~gg™ ¼  „ ¥ ' œ$ ‘5 ¥¦ ©A ˜s«6 Œ ËÊ ™ ¥¼ f ÒÊ q PC „ . the probability of getting is and the probability of getting is .2) æA 6 bB)5$ This determines for every set .CHAPTER 1. Furthermore. Because . we construct the experiment so that the probability of getting is . and if the set has uncountably many points. but not the number . this requires that every number have probabilty zero of being chosen. This distribution was discussed immediately following Definition 2. Introduction to Probability Theory 35 In addition to tossing a coin.11 Suppose we choose a number from in such a way that we are sure to get either . the measure induced by the stock price . when the initial stock price and the probability of is . we define to be the Lebesgue measure of the set.12 Uniform distribution on . define (4. For example. perhaps using a random number generator. Example 1. For each Borel set . to be . 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. then this probability can be positive. the collection of all Borel subsets containined in . Nonetheless. Here are some probability spaces which correspond to different ways of picking a number at random. just as I know of no way to toss a coin infinitely many times.12 provide probability spaces which are often useful approximations to reality. the probability of the interval is . because this interval contains the numbers and . Let and let . Define the standard normal density ‚uDEè7§o2 ¬ ‘@ ¥¦ x ' œ $ ó ¼ H†œ ¥ Ä Let . another common random experiment is to pick a number.8. both Examples 1. 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. We describe this random experiment by taking to be . Example 1.13 Standard normal distribution. Example 1. or .

If is nonnegative but otherwise general. but if a set is given.e. we define is a set in A˜@w¨ ’ ° œ A œ w¨ ’ if if .2) is an interval . possibly also taking the values to be a random variable on this space. we define is simple and for every for every and for every . With this sequence.e.e. and let . then we can write (4. we can define A „ ¨ ’ ‚‚‚A pA©  àA » lss«B7j˜SS8 In fact. and every single point has probability zero of being chosen.14 Let i. i. If is a simple function. i.2). we define such that ç & $ ‚fQ§ p %p # £ xwBYÞ Œ ¥¦ Í x  ß A — © ¢ ó qÍ Ég Ž œ 6  '  ' © ² t—d’$ ¬  ¥ 4—d’$ Ž ßA wBYÞ Œ ¥¼ ' ¥A ™ ¦ ˜€"A „ „$ ÷ Ž Ž Ž h h h œ If in (4. is an indicator. . a mapping from If be a probability space. then the probability the point is in that set is given by (4. The construction of the integral in a general probability space follows the same steps as the construction of Lebesgue integral.36 This corresponds to choosing a point at random on the real line.. we can always construct a sequence of simple functions ‚ ü „ H’ ¨ ' —}’$ Ž ƒ—d’“8 à ' $ ž î ê x ¥ ‚ƒ¦ ¥Q§ƒS8 Ô íyìëb» ¦ X§ Ž Ô » ž ó ó ' $» —d’RY8 uíyë ˜t—d’j8 î» ê  ' $ „ ¨ ’ ‚‚‚ à ' $™8 à ' $q8 à ' $! 8 à siƒ—d’QYTƒ—d’7„ƒ—d’bPTo6 Ô 8 ù Á ¦ ¥X§T8 ó û ú)Bø x  ¦ Q§ Ž Ô ¥ ó ” where each is a real number and each ÷ !P˜” ÷ fB”   !   ‚ œ ¥ l'u”@$¦ s”  ¥ ¦ Q§ ¬ ¥ Ô ” x ¥ v¦ X§ Ž Ô   ó »¢ ó »¢ Ö ‘œ ™ ” ÷ fB” !   A l'—}’$  ' ¬ ¥ ” ¢ t—}’$ Ž   » Ö ‚ œ y'@5$¦ ¥xg¦ ¥X§ Ž Ô  ó ™ ¨ j†œ for some set .2) as the less mysterious Riemann integral: . Definition 1. We repeat this construction below.

Theorem 4. almost surely is enough. then We restate the Lebesgue integral convergence theorem in this more general context. then In fact. Introduction to Probability Theory 37 where then we define The expectation of a random variable The above integral has all the linearity and comparison properties one would expect.e. In particular. if and are disjoint subsets of and is a random variable. we don’t need to have for every in order to reach this conclusion. Ž „ ÷ A ¦ ƒX§ Ž Ô ó A ¥§ ƒ¦ X•8 Ô T¦ X§ Ž Ô n ¥ ó ó ’ ! ‚ƒ¦ ¥Q§ Ž ! I¦ Q§ Ž ¬ ¦ X§ Ž %i¬ n ¥  ¥ ó ó ó  œ ' 8 à ' —d’$“€—d’$ ¨ '—d’$“–—d’Ž $ Ž 8 à ' „ ’ ‚ ¥ ƒ¦ X§•8 Ô Ã¦ X§ Ž Ô ¥ ó ó ¨ H’ ÷ „  ¦ X§ Ž Ô ¥ ó ¦ QY…n Ž $ Ô ¥§' 8 ó ÷ Ž ž î ‚ S» 9 ë b» ê Ž ¥Ó "87 ëy Tà Ž ¥Ó ž A ¥ ƒ¦ X§ » Ž Ô "87Tbíyî » ë T¦ Q§ Ž Ô 9 ë ê à ¥ ó ó ‚ƒ¦ Q§ Ž Ô x  Ž ¥Ó ¥ ó Ž ‚ƒ¦ Q§ Ž Ú ¬ ¥ Ô x¦ ¥Q§ Ž ¬ ¥  ó is defined to be ‚‚‚A pA©  àA lis)˜7–‡˜S» Ž   ó  ’ ' $ 8 à ' —d’“–—d’$ Ž ™ 8 œ If is a set in and Ô ¨ ¨ ¥ Ô x ¥ Ô ‚ƒ¦ X§ ¥ — Ž ó I˜¦ Q§ v Ž ó ¦ Q§ Ž ó AQX6˜l'—d’$ Ž Q…Yuyx€'—}’$ — Ž QX˜l—d’$ Ž )uo€—d’$ v Ž A ¨ {z  A6A' … { z y x ' A˜H¦ X§ — Ž Ô ˜H㦠Q§ v Ž Ô É ¥ A É  ¥ ó ó is a random variable. we say it holds almost surely. Then or equivalently. then If for every . Finally. if and are random variables and is a real constant.4 (Fatou’s Lemma) Let be a sequence of almost surely nonnegative random variables converging almost surely to a random variable . we define If Ž Ž h Ž is integrable.CHAPTER 1. . We acknowledge in these statements that conditions don’t need to hold for every . it is has probability one. i. When a condition holds with enough if the set of for which probability one.

4) (4. Theorem 4. we constructed a probability measure on by integrating the standard normal density. Definition 1.6 (Dominated Convergence Theorem) Let be a sequence of random variables. the market measure and the risk-neutral measures in financial markets are related this way. We say with respect to . Now suppose is a function on . Assume that there exists a random variable such that almost surely for every Then or equivalently. In Example 1. and we write that in (4. Assume that ¼ 2 ' ¥ ÝA ¥ "'¼ $~"G¼ 5$ ž îê ‚ S» Ž ¥Ó íyìëb»  Ž ¥Ó ž Ô A § ƒ¦ ¥X–» Ô bêî » ¥ Ž ó íyë µ¦ Q§ Ž ó ‚ 4à 8Tà » Ž D¦ è ‚ E¦ X§§ W2 ¥  ¦ X§ ¥ ‚ ¥§ ƒ¦ Xñ Dè§ E¦2 ¬ ‘‘5 ¥¦ x ' œ $ õ 7ó ô ' ¥A ' ¥$ ÝA ¼ ¦ ˜l¼ 5"G5$ Ô Ž ó à à Ž ˆ! Ž G6 ¥Ó õ 7ó ô ó ÿ ÿ ¥¦ 8 ñ 2 (4.13. We want to show that ADè§ 2 uE7Eñ ADè§ 2 uE7o„ñ õ ¦ó ô  ¥§ ¦ Xñ õô ó We can also evaluate 2 D Eè ‚ ' ¥$ Ý ¨ l¼ 5~jœ for every ©  Dè§ €„¦2 þ õô ‚ ‚ p  A l‚ssA«BA7©vdàBŽ Y» Ž ž î b» ‚S» Ž ¥Ó íyìëê  Ž ž î º ¥ A § ƒ¦ ¥QX» Ž Ô íyë u» ¦ Q§ ê ó ÚÚÚ Ã is™ Ž ‘q à ‚ almost surely ‚‚‚A pA  àA lsi)B¦© …BYŽ » Ž be a sequence of random . In fact. we call a density and we can define an associated probability measure by We shall often have a situation in which two measure are related by an equation like (4.5) .14 gives us a value for the abstract integral which is an integral with respec to Lebesgue measure over the real line.3). In fact.3) 2 (4.3) is the Radon-Nikodym derivative of The probability measure weights different parts of the real line according to the density . whenever is a nonnegative function defined on satisfying .38 Theorem 4.5 (Monotone Convergence Theorem) Let variables converging almost surely to a random variable Then or equivalently. converging almost surely to a random variable .

We have already proved that ‚Dè§ 2 u„¦Eñ õ 7ó ô  ¥§ ¦ Xñ õ ¦ó ô É ð ”jà We let get and use the Monotone Convergence Theorem on both sides of this equality to ‚ 4à for every A ¥ ¨ G¼ €€Í Dè§ 2» E7)iñ  ¥§ » †¦ Qƒiñ ž õ 7ó ô õ Xó ô ¼ wjÍ ¥ ¨ ' Í$ » î » ê  ' Í$ Y5Psñ uíyë ˜tY5tñ ‚ ‚ ‚ à ' Í$ ™ ñ à ' Í$ q ñ à ' Í$ ! ñ à isƒY5QiHƒS7bIƒS5b¦T6 ñ ñ ‚‚‚A pA©  à iss«B7`BA » ñ ” where each is a real number and each is an indicator function. In is a .for in (4. Introduction to Probability Theory 39   ñ ÷ è D ¦§ .e.5) and an equation which is suggested by the notation introduced in (4. a linear combination of indicator functions..5) holds when is a simple function. We can always construct a sequence of nonnegative simple functions such that and for every .12. Then for every ñ Step 2.e.5) holds when is an indicator function.4) (substitute “cancel” the ). Step 3. Assume that is an indicator function..5) becomes for some Borel set 2 sr §bpp qô 'Í S$ ¬ ¥ €Sxñ  ' Í$ CHAPTER 1. Now that we know that (4. ¼ œ ¥ Ä ‚Dè§ 2 bE¦Eñ  ÷!  )P˜” õ ô ó D„è¦2 v Ys” § ”ö  » ¢ t õô ó ÷ P˜” !  DE7uYö ” è§ 2”  õô ó » ¢ ÷ P˜” !  ¦ XrYö ” ¥§ ”  õ 7ó » ¢ ô ÷!  )P˜” ¦ ¥X§‡”Ys” v ö  ¦ Qñ ¥§ t õô ó » ¢ uõ ô ó ” Yö ÷ !fB”   A Í ö l'Y5$)”Si”  ' Í$ tS5tñ »¢ ‚Dè§ bE¦2 ¬ ñ '‘5$ ¥¦ œ  ' œ$ –‘5 ¥¦ ó Step 1. assume that simple function. The standard machine argument proceeds in four steps. Now that we know that (4. This is true because it is the definition of . (4. In other words. i. we consider a general nonnegative function . i. that case. We include a proof of this because it allows us to illustrate the concept of the standard machine explained in Williams’s book in Section 5. page 5.

we obtain the desired result: 1. To avoid trivialities. In the last step.e.40 Step 4. knowing that does not change the probability you assign to . we have Subtracting these two equations.1) . This discussion is symmetric with respect to and . The probability that is . 1.. and is the outcome. suppose we toss a coin twice. if and are independent and you know that .5 Independence In this section. Whether two sets are independent depends on the probability measure . Conditional on this information. i. with probability for and probability for on each toss. By integrable. which can take both positive and negative values. we consider an integrable function . Suppose you are not told . we define and discuss the notion of independence in a general probability space .1 Independence of sets Suppose a random experiment is conducted.5. we mean that ¿From Step 3. although most of the examples we give will be for coin toss space. For example. the conditional probability you assign to is still the unconditional probability . the probability that is The sets and are independent if and only if this conditional probability is the uncondidtional probability . but you are told that . Then œ ¨ T¸’ ' $ A ¥¦ œ ‚ q ‹ ~e•s…p¦ €A ‹SŠ Wc‘s…p¦ ¥@ewcb…p¦ €A q Š @£dbp ¥¦ e ¥ e ¥   c c… © Š G6  Š e ¥¦ ¨ © Š €v ‹ ' $ pA ¥¦ ‚ pfT5 ¥¦ ‘ 5 ¥¦ '  á œ $ x ' œ $ ÿ  ¨ y’ ’ ‚  ¥ œ$ ¥  '  á œ$ l'x$¦ "'‘5¦ hwT ¥¦ ™ ¨ ‰g ™ ¨ jœ Definition 1. we assume that .15 We say that two sets and ‚Dè§ 2 7E¦Eñ õ uó DE覧2 — ñ ¨ Dè§ ƒE72 v ñ õ 7ó ô õ 7ó ô ¦ X§ — ñ ¥ ¨ ¥§ v ˜¦ QEñ õô ó õô ó ‚˜ÉH¦ X§ ¥ A É  ¥ — ñ õ 7ó ˜H㦠Q§ v ñ õ 7ó ô ô ‚Dè§ u„¦2 — ñ õô ó ADè§ u„¦2 v ñ õô ó are independent if ñ  œ   ’   ¦ Q§ — ñ ¥ õô ó ¦ Q§ v ñ ¥ õô ó  ¨ g’    œ c ¦ Xñ ¥§ õô ó œ €¨ ’ ' œ$ ‘5 ¥¦  ' ¥A ™ ¦ B€wA „ $ œ ' œ$ ‘5 ¥¦ œ j¨ ’ (5.

1).16 Let and be sub. We say that and are independent if every be the -algebra and e e €e  !q  Š c Ü Û )6 )vpA ¥¦ © ‚ 6  ' $ !q c ‹ SŠ  ‹ q Š p ‚ SŠ hfT5 ¥¦ ‹  '  á œ$ A p  ' $ ¥' œ$ ‹ q Š ƒA¦ w@5 ¥¦ A ‹ p  ' $ YŠ ƒA ¥¦ A Š  SŠ n q Š h@5 ¥¦  ' œ$ ‹  Û ‚ 7Û )6 œ ' $ ¥' œ$ ¥  '  á œ$ pA¦ "‘5¦ h‚T5 ¥¦ c ewb††w…œ c…   á  c eA e c…  £@"‘€u† ¥¦  ¡ c ©)6 ‚«6 Š !q ¯ €  e ' $ A ¥¦ ¯ e c  e cA c c…  wBw‡u†˜œ ¯ Let set “one and and one . Let determined by the first toss: contains the sets ‚ ¯ ¨  A  ¨ ƒ€ƒwf†œ for every  ¯ ‚  e eA c e…A  e cA c c… A Q›€1€‘s˜Q‘w"€dbyA „ )š ™ Definition 1.CHAPTER 1. 1. which is the case if and only if . the probability of one and one is the probability of a on the second toss. . in fact. we will find that the product of ¯  e cA c c wBwdb… These two -algebras are independent.-algebras of set in is independent of every set in . conditioned on getting a on the first toss. then . is still . By far the most likely outcome of the two coin tosses is .” Then . Suppose however that . i. In words.14 Toss a coin twice. If you are told that the coin tosses resulted in a head on the first toss.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 compute on the first toss” and is the !q  Š e . are told that the first toss resulted in . Introduction to Probability Theory 41 These sets are independent if and only if . . However. then we have ‚   c c… ¥  c eA c c á  e cA c c q Š @£dbp¦  ¥ £@w€db… rwBw‡u… ¤ ¥¦ A q Š t' SŠ n q Š s' SŠ n q Š g–W‘"w‡up¦ bwBwdbp ¥¦  ‹ $ ‹ $   c eA c c… ¥ e cA c c… ‚  e eA e c…A  c eA c c… A Q›€1€wb˜Qd‘1€dbyA „ )š  ¯   c eA c c £@w€db… ¡ ¡  Let be the -albegra determined by the second toss: contains the sets from ¡ ! ™  ’–¯ Example 1. the probability of one head and one tail. and let be given by (5. is the set “ .5. which is . For example. it becomes very likely that the two tosses result in one head and one tail. If .e. which is now the probability of a on the second toss. and . the probability of . In fact. is . if you the probability of one head and one tail is quite small. if we choose the set the set from .

the sets are indepedent sets. Then . Example 1. which has probability . the intersection of and contains the single element . We say that the different tosses are independent when we construct probabilities this way.17 We say that two random variables and are independent. as shown by the following example.15 Define and for every set .3 Independence of random variables Now the pair pair The set in this last equation is a subset of the plane . the measure induced by ‚ ¥ Ä œ A œ G¼ H†€Q›€¨ Ž p¦ •@5$ … ¥ x ' œ Æ –Å 8 Suppose and on to be are independent random variables.42 Example 1. the price independent of . In particular. In the case of . so the product of the probabilities is . generate  c c £db…  c eA c c £@"wdb… !Ê „ bq  e A c wcBwdcb… Ê Ë ™ €£c@"wcdbp ¥¦  eA c…   e cA c c… “wBwdbp ¥¦  c eA c c  £@w€db… € c bË …  e A c wc˜€dcb…ï !  c b… ¥¦ ©  $ g€' „  ¥¦ œ ' œ$ @ ¥¦ Ä „ œ   c e… ¥   e c… ¥ A ©ƒ ~›€sp¦ €A ©ƒ @W‘sp¦ €A pÛ @›€up¦ @A ©Û @£dbp ¥¦   e e… ¥   c c…  e eA c eA e cA c c… ‘€1w@1@wBwdb† „ ¥¦ for the individual elements of to be c Ž and are independent if the -algebras they 8 q PC ' 8$ ï&–¡ È …l GÇÇ ' Ž –¡ $ § is is Ž å F ¥¼ Ž . È …l GÇÇ ¡ È …y §ÇÇ q PC e Ž 8 Definition 1. and we can define the measure induced by the could be a “rectangle”.14 illustrates the general principle that when the probability for a sequence of tosses is defined to be the product of the probabilities for the individual tosses of the sequence. This is because depends on only the first two coin tosses. depending on whether the third coin toss is or . then every set depending on a particular toss will be independent of every set depending on a different toss. Note that the sets on first toss and on second toss have probabilities and . every set defined in terms of and just considered. whereas either or .5. It is also possible to construct probabilities such that the different tosses are not independent.e. 1. We defined earlier the measure induced by is .17 says that for independent random variables and is independent of every set defined in terms of . a set of the form . These sets are not independent. In this case. and £c€e€1A€cdc€"Awc@ewBAwcdcdcb…€ˆ † e e c  ‡F qPC È …l GÇÇ 8 Definition 1. so is a probability measure. i. Ž  e e cA c e c…  D C§ F  q C €wBw@wbïiEXSPb… ample. where and . AQG‚¿swƒ€¨ …    ‘ œ ¨ ' 8 $  ¨ 8… á  œ Ž †Gw’€ƒ"A Ž %… ¼ Hg ¥ Ä ¼ Hœ ¥ Ä  ‘ “’œ q ¥¼ å ‚ q ¼ HÄ X ƒG"A Ž %p¦ t' $ ¦ ¥ A ¨ ' 8 $… ¥  å å å ‰ –Å Æ ' 8 "A Ž $ q ¥¼ ‚ ¥ Ä  A   ¨ 8 … ¥ x '  G¼ HvQ‚wUsp¦ ‘x$ takes values in the plane 8QÅ ‰ Similarly. On the other hand. define to be the sum of the probabilities of the elements in . for ex- p of the stock at time In the probability space of three independent coin tosses.

 ¶™ ' Ž h¡ q Ž $ ç ' 8$ ï0–¡ ' d$ ‘A–¡ ' Ž –¡ $ ' ™$ x–¡ ' Ž –¡ $ ™ ' Ž –¡ $ ' ™$ †A–¡ Ž Ž Ž A  œ ¨ '' Q›€t"—d’$ Ž rifl†€h—d’v™ Á l… $ òe ’…   œ ¨ ' $ ’ ' d$ @Ah¡ ' –¡ $ ' ™$ †A–¡ Ž ' 8$ ö  04d P ROOF : Let us denote and is of the form a typical set in . A joint density for is a nonnegative function Not every pair of random variables has a joint density. Since every set in is independent of every set in . and differentiate with respect to both and . for independent random variables and . Then and are independent variables if and only if the joint density is the product of the marginal densities. if . Let to . which means that will contain all the sets in and others besides. which means that contains at least as much information as . this is the case. can contain strictly more information than . for example. This follows from the fact that (5. the joint distribution represented by the measure factors into the product of the marginal distributions represented by the measures and . (In general. Introduction to Probability Theory 43 In other words. Therefore. we can show that every set in is also in . we have that every set in is also in .CHAPTER 1.   ˜€1g A É ¨$   Í  ÍA É ¨$  X˜˜€1gHœ 8 ' Ž fg ™ $ ò ‰ cÅ Æ –Å These have the properties 8 Ž ‚ ¥ Ä  § G¼ Iv`A  Y'  $ ñ ! ÙA$  '  ó A ¥ Ä œ Aͧ'Í G¼ H†ƒPXSY5$ ‰ ñ ¬  ‘5$ ' œ ó Æ ž ‚S–˜Y'  «—–$ ¦ ñ ž — '  $ ºS•S)P5$ ¦ ñ § A ñ A ”§ ' ”A Í ó ‰ Æ ‰ ‰ Æ ' 8 wA Ž $ ‚  @XS'  f5$ ¦ ñ ! ¬ hw’5$ § ͧ A Í  '  ‘ œ ó ó '  P5$ ‰ ¦ Æ ñ AÍ ‰ Æ 8 Ž ‚'  ' œ lpA$ ‘5$    ¨ 8 … ¥  ‰€ … h Gwvsp¦ ucÅ ¨ Ž p ¥¦ Å  œ Æ ¥ G‚¿s… –€¨ iw ¥¦   ¨ 8 á  œ Ž…¤ such that Ù'p“‘’5$ ¦ œ ‰ –Å Æ ž ž— ' ™$ †A–¡ ' 8$ ï&–¡ ' ™$ †A–¡ 8 Ž ó ' 8$ ï04ö ¦ ‰ –Å "A $ Æ ' 8 Ž  'Í tY5$ ñ Ž Æ ' ™$ †A–¡ ' Ž fò $ 8 8 ' d$ ‘A–¡ ™ ¦ cÅ ‰ ‰ Æ Å 8 and Ž Ž ¥¼ Ž and are independent if and only if (5.) In the same way that we just argued that every set in is also in . but if a pair does.2) is and . we conclude that every set in is independent of every set in .7 Suppose and are independent random variables. Then and are also independent random variables.1) in terms equivalent to independence of and . which is defined in terms of the random variable . this set is in . In fact. We must consider sets in ö ò Theorem 5.2) Æ –Å ¥¼ and be functions from and . write (5. But . Take of densities. then the random variables and have marginal densities defined by Suppose and have a joint density.

Unfortunately. 1. Consider the following example. If both have positive variances. so would be and . If almost surely. the correlation coefficient is zero. but be independent of and have .4 Correlation and independence provided all the expectations are defined. 8 According to Theorem 5. are independent. and if and ‚‚‚A $ ¡ sisl' q Ž –w¨ q Bl' ! Ž –w¨ ! œ œA $ ¡ ‚ ' » œ$ ¥Ú Ú Ú ' q œ$ ¥' ! œ$ ¥  ' » œÚ Ú Ú á q œ á ! œ$ l)€5¦ ÂsiPl‘5¦ "l5¦ tQ•(isdt‘€r›5 ¥¦ Ž are functions from ‚‚‚A A ssisq Ž l! Ž 'S5$ ¬ ¥ pS5rò Í  ' Í$   Ž ò à 8 q ˜ q Ž ¥¼ ¥¼ .8. 8 q 8 q Ž 8 d Ž Ž 8 d Example 1.5.16 Let be a standard normal random variable.8 If two random variables to . and are uncorrelated. and were independent. Define standard normal random variable.18 Let variables are independent if for every sequence of sets and for every positive integer . for independent random variables. we define their correlation coefficient and ç which is true because and general functions and . Now use the standard machine to get the result for Ž Ž ö ‚ 8 ¥ —pÓ iÚ ¥ ¨  Œ Ž Ó W„‘8 Ž ¥Ó  i' 8 ¥ $ ¥ SïpÓ …¨ &8s' Ž Ó q¨ Ž &$h ¥Ó "A Ž $ g x ' 8 8 Ž ‚ q  Ž Ó …¨ Ž Œ Ó ƒ' Ž $ ¥ ¥ x Ž ö 8 Ž A  ¨ 8… ¥ œ Q‚wUsp¦ b›j¨ Ž p¦  ¥ Gwvsw—›€¨ Ž sw ¥¦ … ¥   ¨ 8… á  œ …¤ ò A 8 l'ï&$röGÓ ¥sE' Ž r›Ó „ˆ@&r"' Ž rò Œ ¥Ó Ú $ ò ¥  ' 8$ ö $ ' 8 ‚ ï0$ ' 8 ' $ "A Ž $ Ž 8 Ž 8 6  © ¨ –€uk kp ¥¦ ju© Ukp ¥¦ d …Ž    d … ' $  $ ! ¥ '  4ö   x ' 8 $ j €"A Ž P 8 Theorem 5. let the distribution . P ROOF : Let trying to prove becomes and be indicator functions. two random variables can have zero correlation and still not be independent. We say that these random Definition 1. but in fact.44 be a sequence of random variables. The last claim is easy to see. Then the equation we are The variance of a random variable is defined to be Var The covariance of two random variables Cov and is defined to be Cov Var Var For independent random variables. then and are independent. We show that is also a and are not independent. the covariance is zero.

presented in the lecture dated October 19. Introduction to Probability Theory We next check that is standard normal. both Cov and Where in We conclude this section with the observation that for independent random variables.e. we have 45 Since is standard normal.5. does the measure put its mass. We now return to property (k) for conditional expectations. of their sum is the sum of their variances. 1995.5 Independence and conditional expectation. if and are independent and then This argument extends to any finite number of random variables. Ž Ž  ‚ Ž Ó SG Ž Œ ¥Ó ¥   ÿ ¡ (k) If a random variable ' 8 wA Ž $ ¦ q ¥¼ ‚6  6Ú ©  d ¥ «l€˜¡Ó lÚ q Ž Ó „ƒd q Ž Œ Ó „‘8 ‰Ž Œ Æ Ó twA Ž $ ¥  ¥  ¥Å  ' 8 8 Ž 8   à Ž …p¦ ¥~  ¿ip ¥¦  à 8…   Ã Ž p¦ @  à Ž p ¥¦ ¨ … ¥  … ‚ Q  à Ž Qp ¥¦ ©p —  à Ž p ¥¦ ©p  ¨… n … uã ˜…p¦ b  à Qp¦ ¥I|¾˜kp¦ b  à p ¥¦  © ¨ d ¥ ¨… n ©  d… ¥ …  à Ž … n ©   à Ž…  ¨  u©˜d Ž p ¥¦  Ž ¨Qp¦ ¥H–u˜d  ÃUs…p¦ Tn—u©˜d  Usp ¥¦    Usp ¥¦ © ¨  u˜d 8 ¥  à 8… à 8… 8 ¼ w¨  ¥ and and and and . page 88. i.3) . then Var Var Var Var 1. Indeed. which shows that is also standard normal. the same as the best estimate of based on no information. For . If we are given independent random variables .. the variance . Therefore. then the best estimate of based on the information in is .CHAPTER 1. we shall need only the second assertion of the property: The point of this statement is that if is independent of . and we have have expected value zero. The property as stated there is taken from Williams’s book. then ‚' lX» Ž $ nÚÚÚ n ' Bsi˜P˜q Ž $ n' „"! Ž $ –Q» Ž nBsÚi4q Ž P! Ž $ ' Ú Ú n n » Ž "issyq Ž l! Ž A‚‚‚A A Ž Ž$ Ž$ Ž $ g ¥Ó ¤ Ž W ¥Ó Ž ¥Ó Ž     x ' d ‘A$ Var Var Var 8 In Ž gd  8 Ž i q @¡Ó WIA5y ¥Ó ' d ¥ ¨ d$g  Ž . Being standard normal. (5. what is the distribution of ? Var Var is independent of a -algebra ‚' 8 lï0$ n S' Œ ¥p n ' G80$ n 8 ¥ )pÓ W¨˜8 Œ Ó & ¥ ¥ Ž Ó q¨ Ž Ó ¦HS' $ ¥ $p ¥ i q 'ï8pÓ ¥W¨†80nP'G8pÓ W¨08$s' Ž Ó ¥W¨ Ž STn q ' Ž Ó W¨ ' ¥ ¨ 8 i q ï8pÓ ¥W¨ Ž Ó …n .

46 The left-hand side of this equation is . then the partial  To show this equality. be a sequence of independent.6 Law of Large Numbers There are two fundamental theorems about sequences of independent random variables. The argument proceeds in two for every . We next check that Var as . We will use this argument later when developing stochastic calculus. Here is the first one. We must for every  Ž Ž . 1. We first check that other words. the random variables are increasingly tightly distributed around as . and the right hand side is The partial averaging equation holds because and are independent. we observe first that also check the partial averaging property ¦ Q§ Ž ¬ ¦ X§ Ž ¥Ó ¬ ¥  ¥ ó ó Ž ¥Ó is -measurable. each with expected value and variance .9 (Law of Large Numbers) Let distributed random variables. In steps. but here is an argument which makes it plausible. The partial averaging equation for general independent of follows by the standard machine. we first recall from (5. Var Var É ð ‘jà As . For the first step. we have Var . identically Theorem 5. Define the sequence of averages We are not going to give the proof of this theorem. since it is not random.5. Therefore. É ð ‘jà è 6 ð qà ‘' » &$ 6 ð 8 » Œ  »8 o ‚ è p è n Ú Ú Ú  n è n Œm   ¥ n Ú Ú Ú n t˜ S…˜sin Tè d© à „i» Ž Ó IBis4q Ž Ó TP! Ž ¥Ó ‘© à GSp ¥Ó ¥ n !   f˜” ˜à ‚ à  qà  ® ” Ž­ q ¡ q ¡ »¢ à 6 ð '»8 ‘t)S&$ !f˜”    '»8 t)S0$ »¢ »S8 è  »  ¥Ó 8 è »8 Then converges to almost surely as . which by assumption must be independent of averaging equation we must check is q¡ É ð ”jà à ‚"l‚isA)˜A7©jà€A ‚ ‚ p nÚÚÚ » Ž Bis˜n q Ž n ! Ž x  » 8 è ‚‚‚A A sislq Ž s! Ž   œ ‚ l'fáT5¦ ¦ X§ %i¬ ¥ Ô ¦ Q§ ! ¥ ¬ ¥ Ô œ$ ¥  ¥ ! l  ¥   ó     ó ' $ ¥' œ$ A¦ "@ ¥¦ ‚ƒ¦ X§ ¥  ¥ § ' $ ! ¥ ¬ †¦ XSx ¥¦ ¬   ó ó ‚  ¨ €fœ .3) that the variance of the sum of independent random variables is the sum of their variances.  If is an indicator of some set . we simply compute times For the second step.

the distributions of all the random variables have the same degree of tightness. ž & $ îê ‚Í PX§ ) — ¬ %p © # ¡ ~›€¨ » kp ¥¦ íyìëb»   œ d… ó s¥ q ¡ È eÈ ¼ Hrœ È » |d q 6 » td !  P˜” !   fB”  '» d t%|A$ ‚ q  à ¡ ¡  ® dà €”# ­ è ¨ Ž q »¢ »¢ » d then each has expected value zero and Var Var q ¡ è à A Pd¨ » lsi˜PPW¨ q PP£¨ ! $ “|d x » # $ n'è 'è Ž$ n Ú Ú Ú n ' è Ž Ž » Y8 à à # » S8 É ð ‡à É ð ‘jà .5. we should divide by rather than . the distribution of approaches that of a normal random variable with mean (expected value) zero and variance . if we again have a sequence of independent. but now we set As . around their expected value .7 Central Limit Theorem The Law of Large Numbers is a bit boring because the limit is nonrandom. If we want to prevent this. The Central Limit Theorem asserts that as . This is because the denominator in the definition of is so large that the variance of converges to zero. In particular.CHAPTER 1. Introduction to Probability Theory 47 1. as measured by their variance. each with expected value and variance . for every set . In other words. identically distributed random variables.

48 .

e. the stock price either goes up by a factor of or down by a factor of . c F the stock price moves up by a factor of if the coin comes out heads ( ). Note We write depends only on . Each is an -measurable function .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 .Chapter 2 Conditional Expectation Please see Hull’s book (Section 9. At each time step. after tosses) under the outcome .6. that Each is a random variable defined on the set . and F D EC § ™…ƒÝ ð ' ™ ïwA „ $ ”C h h ! E— ” ¡C .) 2. that is.1) The collection of all possible outcomes (i. Consider a sequence of 3 tosses of the coin (See Fig. Then is a -algebra and is a measurable space. It will be useful to visualize tossing a coin at each time step. We will see later that R is in fact 49 ’ ™ ’ ” PC ¼ Hð ¥ ” ’A ‚‚‚Aq ’A! –1Sisis–1s€’ ' $” —d’QPC “ “ ' $” —d’QPC “ ” —’ ’ „ ‚  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€ƒ"w@ƒ"€w€"‰‡‘1€d€"€€‰Bw€€"€wdBwddb† „ '„ „ “vV™ $ u  ¡ Ý ™ ” PC „ ‚ ' q ’ A ! $ q C t ' ™ ’ A q ’ A ! $ q C t ' $ q l˜|es€d’bflxeu|es€d’uP–d’7PC A ' ! $ ! C t ' ™ ’ A q ’ A ! $ ! C t ' $ ! l"€d’yxl–1i|es€d’s4–d’b4C e § down by a factor of if it comes out tails ( ).e. and will denote the th element in the sequence . More precisely. is a function where is the Borel -algebra on I . Thus in the 3-coin-toss example we write for instance. 2. and say that Note that we are not specifying the probability of heads here. to denote the stock price at “time” (i. sequences of tosses of length 3) is A typical sequence of will be denoted . let .

1 (Sets determined by the first tosses.50 ω = Η 3 2 S2 (HH) =u S0 ω =Η 2 S (H) = uS0 1 ω1= Η S0 ω1 = Τ S (T) = dS0 1 ω =Τ 2 2 S2 (TT) = d S0 ω3 = Τ S3 (TTT) = d 3 S 0 ω = Η 3 ω =Η 2 ω =Τ 2 S2 (HT) = ud S 0 S2 (TH) = ud S 0 ω = Η 3 ω = Τ 3 ω = Τ 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. for each . are defined similarly. z {y Example 2. “ ¥¼  The sets we define etc.1 . ” ™ “ .-algebra of .2 Information Definition 2. R R For any random variable defined on a sample space and any . It is easy to check that is a -algebra.1: A three coin period binomial model. Similarly for any subset of Ý q „ œ ¡ Ý ‚  ¨ ' XGwƒx}’$ Ž „ Hl†xGw¨ Á ¨ ’ … t   A X  ‚ à ' Q  ƒx}’$ Ž „ Hl†w  à Á ¨ ’ … t ¼ €¨  ¥ „ “ “ ¡ ” ™ ¡ ™ 6 t § t F Ž…  Ž yQ  r Ž lQ   Ž … …A …A Ž… œ Ž ¡ . we can decide whether the outcome of all tosses is in . knowing only the outcome of the first tosses. Recall that the Borel -algebra is the -algebra generated by the open intervals of I . the collection of sets determined by the first toss consists of: àA‚‚‚A pA©  BXsis)˜7G“ ” ™ ” PC Note that the random variable is -measurable. we will use the notation: Assumption 2. 2.) We say that a set is determined by the first coin tosses if. measurable under a sub. In general we denote the collection of sets determined by the first tosses by . In this course we will always deal with subsets of I that belong to .1 In the 3 coin-toss example.

) Let be a random variable . The collection of susbets of determined by is a -algebra. . Conditional Expectation 1.CHAPTER 2. The complements of the above sets.2 (Information carried by a random variable. then is generated by the collec- . 6. Another way of saying this is that for every . Any union of the above sets. is given by consists of the following sets: '–d’$ ¼ Tð Ž „ ¥ ' Ž t¡ $ „ ' Ž t¡ $ Ž Ž Ž œ w¨ ’ ¡ ˜  œ á ! rwP'  $ „— Ž ‹ ˆ ˆ ˆ ‡ ƒ ˆ ˆ  ~€ Œ Œ 'VVe†QV–h••V| ‹ ˆ ƒ ˆ‡ ƒ ƒ ˆ } '¡Q……Q~€ ”Œ | ‹ ˆ ˆ ƒ‡ ƒ ˆ ƒ Œ } 'V¡S…Q¡G~€ “Q| ‹ ˆ ƒ ƒ‡ ƒ ƒ ƒ } } '‰…W†……G~€ ’Q|  ‘y  Ž ‹ ˆ ˆ ˆ‡ ƒ ˆ ˆ‡ ˆ ƒ ˆ‡ ƒ ƒ ˆ Œ 'VV…QVV¡Q……Q~€ V| ‹ ˆ ˆ ƒ ‡ ƒ ˆ ƒ ‡ ˆ ƒ ƒ ‡ ƒ ƒ ƒ  ~€ } 'VŠS…Q‰WV¡…S…†…„‚Q| .2 (Sets determined by ) The -algebra generated by ‚Q€j Á A$ E— Ý ¨ '  ! ' $ Ž h¡ ' Ž –¡ $ ¨ '' Á  „ H’ "|}’$ ÿ  $ Ž …†r' Ž h¡ ¼ ¥Hð „ ¡ ! Ž $ E— Ž … If the random variable tion of sets takes finitely many different values. . 4. we can decide whether or not . 7. 7. . knowing only the value We say that a set of the random variable. Complements of the above sets. 5. and denote by . 3. 5.  ”™ ‡ £ ¤¢ ˜‹ ”™ ©h€  ™ h€ ¨§¦‰“Q|  }Œ | ¥ Œ} £ ‡˜‹ ”™  ¤h€  ™ hQ'VVe†QV–h€ •V|  € ‹ ˆ ˆ ˆ‡ ƒ ˆ ˆ Œ Œ £ Ÿ ™ ‹ ”™  ¢h€ˆ ¡œ ¥ Wž' “5h›'‰…W†……G€ ’Q| œ € ‹ ˆ ƒ ƒ‡ ƒ ƒ ƒ } }  ”™ š Example 2. 4. Any union of the above sets (including the complements). 2. The collection 1. . 2. . 3. and . . 2. . is determined by the random variable if. . . of sets determined by the first two tosses consists of: œ q—'  $ !E— Ž q „ rœ Ž ‹—« …’¡œ Ÿ ”™ ‚€  ª     ‹ Ž ’¡œ Ÿ ”™ ‚€ Ž     Ž  ¡ ¼ ‡¨  Ž ¥ Ž . 51 . these sets are called the atoms of the -algebra In general. either or . 3. 4. if is a random variable . which we call the -algebra generated by . then 1. 6. Definition 2.

then even without knowing . If we know that . e. From elementary probability.3 Conditional Expectation In order to talk about conditional expectation.g.52 2.3. then even without knowing . given . then 'q ! C$ BPC 4“ ¥Ó ÿ 2. is a random variable whose value at is defined by – If or . i. then . . . then the value of ÿ ’ e e e €•† ’ ' q ! C$ lfC 45j ¥Ó h ÿ ' C –d’$ q   where . Properties of : should depend on .3 (Expectation.e. should also be known. . so that. we know that .. ..) If then and We can think of Let us estimate .1 An example œ ' Ž ¬ 5j ¥Ó ¥$ Ž ¬ bª ’ ‚ l'—d’$¦ "'|}’$ ¥  ¥ ¬  ¥$ Ž ¢ Ù¦ Q§ Ž ó d' Ž ¬ 5j ¥Ó œw¯’ 6 t ' ¨ œ w¨ ’ © ² –d’$ ¬ ¥ as a partial average of A ‹ q Š t@w‡ ¥¦  ' e c c$ ª ’ ‚' $ ¥' l—}’¦ "–d’$ Ž „ t Ž ¥Ó ¢ ' q C ! 5j ¥Ó C$ A l'  –qPC 45jÓ ––d’€8 ! C$ ¥  ' $ÿ ÿ ’ 8 e c is the probability of q œ ' $ ¬ ’ ¬ t ' œ $ „ r„® —d’ ¥¦ uª ­‘5 ¥¦ h q PC „ ! 4C ' Š €egt Ih ¨ ©$ ‹ ' © A 6$ Š si)5q¨ £h 'q ! C$ ˜PC 45j ¥Ó . Definition 2. q „ ­œ ÿ h h etc. Denote the estimate by . We define – If or . the coin tosses are independent. If we know that . we know that . we need to introduce a probability measure on our coin-toss sample space . In particular. Let us define is the probability of . We define D EC q D EC q ‚ C ' c e e$ ' q ! C$ ¥  ' e e e$ ' q ! C$ 7DEX§ td@•0llPC 4“Ó t@€€&sBfC x5j ¥Ó ÿ ’ DC§  ' $! EXÿ t—d’bxC § ' q ——}’$bPC D §  ' $q RC q gƒ—d’uPC c e e  ‘€H’ ‚bEYF4'@ewc‡csBfC x5jÓ Pddd5s˜PC 45j ¥Ó DC  $ 'q ! C$ ¥  ' c c c$ 'q ! C$ ÿ ’ DRCSt—d’bxC F ÿ ' $! Ft'—d’uPC  $q D F  ' $q RC q t—d’7fC e c c  w‡£’ ' q ! C$ lPC 45j ¥Ó cddq’ c c  q PC If the value of is known. it is a random variable. if if over the set .

We also write 53 ¯ 8 .-algebra of . we can write Let on Please see Williams. is random only through dependence on . In other words. be a probability space. we define . We then take a weighted  ' $q €—d’7fC ‚ SŠ  q YŠ n ‹ q Š n q YŠ n ‹ q Š –‘5 ¥¦  ' œ$ ‹ p ‹ ‹ C§ CF C§ D X ! C D S† ! C D XSF DC§F  ' $q EXS€—}’7PC  e c eA c c eA e e cA c e c…  œ ¨ w€"jd@1‘€wBw@wbv†q’ . and let be a sub. The random variable Then Furthermore. then or .83. Conditional Expectation – If . has two fundamental properties: if if if 2. If we know .Ž ÿ „$ „$ A '' $ q C$ ò  ' $ 'q ! C$ lw—d’sP5rIh|d’iBPC 4“ ¥Ó ÿ ‚ ¥ §! ƒ¦ X"4C ¬ ¦ X1lPC x5j ¥Ó ¬  ¥ §' q ! C$ ó ó ÿ ' $ ‘œ   ' $ ' q ! C $ ‚D C'§ n F bEwQTp5$ !q  ¦ QB54G¥¦ ¬ þ ––d’s˜PC 45j ¥Ó ¥§! C ÿ   D C' § n F ‹ E"XT$ YŠ DC ‹ n DC EQ§ q SŠ tEX§ ‹ q Š –ESF q SŠ tEYF ‹ q Š n DC ‹ n DC ¦ XB4C ¬ ¥§! ó CHAPTER 2. p. Then is defined to be any random variable that satisfies: ™ 8 ¡ ¯ 'P¯ Ž j ¥Ó $ ' ¥A ™ ¦ Bj"A ' ¥A ™ ¦ ˜j"A (a) is -measurable. Let be a random variable .2 Definition of Conditional Expectation where is the function defined above.3. then we do not know whether average: For œ €¨ ’ where ‚ƒ¦ XB4C ¬ Ù¦ X"˜PC 45j ¥Ó ¬ ¥§!  ¥ §'q ! C$ ó ó 'q C$ ¡ ¨ ˜P5–j†œ ÿ h ' q 5–¡ ' q C ! 5j ¥Ó h C$ C$ ÿ 'q ! C$ ˜PC 45j ¥Ó ò ÿ A ' Í$  ' Í  q ! C$ lY5fò tSƒPC 45j ¥Ó ÿ qPC 'q ! C$ ˜PC 45j ¥Ó ÏÑ ÿ D §  EC Í DC EQ§ C §q F  C' § n F  ' Í$ D QYÍ D "XT5$ !q ÎÏ 4Sfò D F  EC q Í DC ESF Ð For every set is -measurable. In conclusion.

In fact the following holds: Œ ¥  ' $ Œ ‚ 1 Ž ‚ ¬ ¥ Ó „ˆE¯ Ž j ‚¥Ó ¬ ¥ ¥Ó ÿ We can rewrite this as œ ’ '–d’$e•8 Ž Œ ¥Ó  ÿ ¯ œ '–’de‘8 Ž Œ ¥Ó $ ' |d’$ Ž ’ ÿ ‚7©ã@˜'|}’$ 3 8˜–'–d’$j8 Á ¨ ’ …  „8 ˜lp ¥¦ Uniqueness. Notation 2. we compute an estimate of depends on the partial information we have about . ’ ' |}’$ Ž ’ œ ‚ƒ¯€¨œ„°Aƒ¦ X§ ¬  ¥ § ' $ ¬ ® ¥ Ž ó ¦ Q"E¯ Ž j ¥Ó ó ÿ „ œ ‚ƒ¦ Q§ ¬  ¥ § ¬ ¥ Ž ó ¦ X8 ó ' $ E¯ Ž j ¥Ó ÿ ‚l"'ï&$P¡ $ ¥ ' $ ' 8 Ž jÓ t 8 Ž j ¥Ó ÿ ’ ’ ’ ÿ Ž ’ 8 satisfying the above properties. which is the intersection of all sets in containing .3 Further discussion of Partial Averaging The partial averaging property is ‚ ( Œ ¥   ' $ Œ Ž ‚ u Ó „ˆE¯ Ž “ ‚¥Ó u ¥Ó ÿ ¯ Lemma 3. i. We then define to be the average (with respect to ) value of over this set . 2. (3.e.3..2) . for all in this set . conditional expectations always exist.10 If is any -measurable random variable.1 For random variables . Thus.1) (3. The value of is partially but not fully revealed to us. is a function of . then almost surely.e. i. . If the -algebra contains finitely many sets. but not told which element of it is. There is always a random variable satisfying the above properties (provided that ). but if ¯ ¥¦ 8 "A Ž ’ œ 8  3 g¸8 ¯ ¨ wœ ¯ ’ (b) For every set . i.3) .e. and thus we cannot compute the exact value of . although the dependence on is often not shown explicitly. There can be more than one random variable is another one. an element of is selected. there will be a “smallest” set in containing .e..54 Existence. we have the “partial averaging property” ' $ |d’"€8 Ž Œ ¥Ó u ¡ ÿ ¬¥ É H Ž ¥Ó ÿ ÿ ’ h h 3 8 (3.. The way is partially revealed to us is that we are told it is in . Because this estimate Based on what we know about . it depends on . will be the same. it is standard notation to write Here are some useful ways to think about : A random experiment is performed.. then provided ÉH E¯ $ ' Ž j ‚¥Ó u ¥Ó ÿ ÿ ÿ ¯ Note that is a -measurable random variable. i.

(c) (Linearity) Proof: Take . And since is -measurable. is of the form .3) holds for and it must hold for as well. Based on this lemma. we have (3.3) for . This set is in since is -measurable. Williams calls this argument the “standard machine” (p. is sufficient to determine .2) by: 2. first use (3. 6  ' œ$ ý`@ ¥¦ ¯ 'P¯ Ž j ¥Ó $ ÿ ¯ ¦ X§ ¥ k Q P $ ¥ ' Ž ¬ þ %H¦ ƒ§w'— ¯ ' Ž j ¥Ó $ ¬ þ ¨ ’ …  6 ÿ$d’lE¯ Ž j ¥Ó „ y†˜œ Á ÿ ‚6 r ' $ «HƒP¯ Ž j ¥Ó ÿ (d) (Positivity) If almost surely. Partial averaging implies . the general measurable random variable can be written as the difference of two nonnegative random-variables . and since (3. unless .3) for each and then pass to the limit. then the best estimate of based Ž Ž Ž 8 Ž ‚ Ž tP¯ Ž j ¥Ó  ' $ 6 Hr Ž Ž 6  ' œ$ –‘5 ¥¦ Ž ¯ (b) If is -measurable. Therefore.4 Properties of Conditional Expectation Please see Willams p. Finally. Next consider the case that is a nonnegative -measurable random variable. to obtain (3. Proof: Just take in the partial averaging property to be . . then Ž „ (a) ‚ ( Ž ‚ u Œ Ó „(„¯ Ž j ‚¥Ó u Œ ¥Ó ¥  ' $ ÿ ÿ (b’) For every -measurable random-variable .2) and linearity of expectations to prove (3.e. The right-hand side is greater than or equal to zero.. satisfies the requirement (a) of a conditional expectation as well. where each is in and each is constant. Such a of simple random variables . i. ¯ ¯ ” œ u u u !   ¬ ± i¬ ¥ ” f˜”» ­ u ¯ ÷ »u — u u u v u u Ž u »u u œ ‚ l' $ ¥  ' ' $ ¥ $ Ž jÓ t"E¯ Ž jÓ “ ¥Ó ÿ ¯ — u ¨ v u  u ¯ ¯ Ž ” ÷ ¯ . then ‚l4¯ q $ Þ n ' ' $ Þ  ' Þ Þ$ Ž j ¥Ó q IP4¯ ! Ž j ¥Ó ! hr¯ q Ž q Hn ! Ž ! 5j ¥Ó ÿ ÿ ÿ ¯ If the information content of on is itself.4) . 56). Conditional Expectation 55 Proof: To see this. we can replace the second condition in the definition of a conditional expectation (Section 2. we write (3.3) when is a simple -measurable random variable. is thus an unbiased estimator of the random variable The conditional expectation of Proof: The partial averaging property holds trivially when is replaced by .3.CHAPTER 2. 88.3. Proof sketches of some of the properties are provided below. can be written as the limit of an increasing sequence but is not necessarily simple. and the left-hand side is strictly negative. using the Monotone Convergence Theorem (See Williams).

-algebra of means that contains more information than .56 (h) (Jensen’s Inequality) If Recall the usual Jensen’s Inequality: is a sub. ‚ƒ¦ X§ Ž d ¬  ¥  Ž ‘d ¬ ¥ Œ ¥Ó ó  ‚ Œ ' ˆP¯ $ Ž jÓ ¥‘d ¬ ¥ ¥Ó  ÿ G ‚8 ¬ 5$j ¥Ó  ¦ Q8 ¬ ' ¥ ¥§ ó 8 8 ' $ ‚ d  E¯ Ž j ¥Ó ‘²U8 ¯€¨œ„°Aƒ¦ X§ ÿ þ  ¥ § ® ¥ Ž d ¬ ¦ X‚8 ¬ þ8 ¯ (a) (b) is -measurable. . then . Then satisfies (a) (a product of -measurable random variables is also satisfies property (b). the -measurable random variable ‚' $ ‚ d  ' y„¯ Ž “ ¥Ó ‘­–P¯ Ž Aj ¥Ó d$ ÿ ÿ ¯ ¯ d (j) (Taking out what is known) If is -measurable. then ‚ ' $ ¥$ ˜ r $ ˜ y"' Ž jÓ 5tfh' Ž x¡ ¥Ó ‚ '' $ ¥$ ˜ r ' lw„¯ Ž “Ó 5tfƒP¯ ' Ž tAj ¥Ó $ ˜$ ÿ ÉH ÿ ' $ ¼ ð ¼e TXw˜ Ž t˜ ¥Ó ÿ is convex and ÿ . then d ¬  u ¥ ‚ l'P¯ $ ¥  ' '  A ¯ $ $ Ž jÓ twbƒ5P¡ Ž “ ¥Ó ÿ' ¯A $ ¡$ ÿ E˜l' Ž t5–¡ ¯ Take -measurable). ‚ $ ¥  ' l' Ž jÓ h Ž j ¥Ó $ ÿ  In particular. then nothing is gained by including the information content in the estimation of . . if is independent of  (k) (Role of Independence) If is independent of . If we estimate based on the information in . as we can check below: 8 Ž ¯ Ž d Proof: Let only if be a -measurable random variable. then acts like a constant.-algebra of . if and Ž Ž  ‚ l' Ž $jÓ ¥t'` P¯ Ž jÓ “ ¥Ó ' $ ¥$ ÿ ¯ÿ ¡ ÿ ¯ ¯ ¯ ¯    (i) (Tower Property) If is a sub. A random variable is 'f¯ Ž Aj ¥Ó d$ ÿ d When conditioning on . and then estimate the estimator based on the smaller amount of information in . ((b’) with ¯ Ž   If of is independent of and . then ¡ ¯  . then we get the same result as if we had estimated directly based on the information in .

for instance. ¥Ó  ÿ ÿ ‚ixC q X§ ‹ ! ' ' q $ ' "!™ PC5jÓ ¥"X§ ‹  q ™ ÿ q w'Q§ ‹ ‘F C n ÿ nF nF Œ Š$ Š$ Š$  ý ! ™ ' q ™ ™ 5j ¥Ó Œ ¥Ó C$ ÿ ‚ ' $ q C' n $  ' $ 'q ™ C$ l—d’uP"X§ ‹ F Š gt—d’s˜™ R5j ¥Ó A similar argument one time step later shows that ÿ   ‚ï·•œ€•)°l—d’s"`™ P5j ¥Ó Š ¨ ’ ®A ' $ '! q C$ ' $ |d’i' ! ™ q 5j ¥Ó lÿ ' · 5 ¥¦ C$ ‚ œ$ ÿ ¦ X(' ! ™ q 5j ¥Ó ³ ¬ ¥§ C$ ÿ ó A ¥§q ƒ¦ QlPC ³ ¬ must be constant on and ‚ƒ¦ ¥Q§˜PC ´ ¬ ¦ X""™ P“ ¥Ó i¬ q ¥ §'! q C$ ´ ó ó ÿ  ¥ §'! q C$ ³ ¦ X""™ P“ ¥Ó i¬ ó ÿ .3. . Conditional Expectation Recall that ¹ •œ · •œ '! q C$ B`™ P5j ¥Ó On the other hand. 57 . Notice that 2. (linearity) . ‚ ¨ ’ ®A' $ ! C' n $  ' $'! q C$ „ •)°l—d’s4"X§ ‹ F Š t—d’lB™ P“ ¥Ó ÿ ‚ ¹ œ ¨ ’ ®A ' $ ! C' n $  ' $ '! q C$ –€w)°l–d’yx"Q§ ‹ F Š gh|d’iB™ P5j ¥Ó ÿ ·•€H)Sl–d’s4"X§ ‹ pF Š $  œ ¨ ’ ®A' $! C' n D C F' EY"X§ n ‹ pF Š $  C§  ' $ C$ D QYF ‹ n D C q F Š ’—d’l' ! ™ q “ ¥Ó ÿ ‚G·€œ€H)°bEXSF n D ¨ ’ ®AD C §  ' $ '! q C$ ‹ –EC q F Š t—d’lB`™ P5j ¥Ó ÿ ³ ‚uDECXSF §  ¥ § q i¬ ‹ n D SŠ –EC q F q Š ¦ XlPC ó This final expression is Similarly.5 Examples from the Binomial Model ÿ ó '! q C$ B`™ f5j ¥Ó  „ t•B•B4k†—`™ A ¹ œ A · œ A ˜ …ÿ  ! CHAPTER 2. We can also write Therefore. We can verify the Tower Property. We compute Now since '! ™ C$ "™ E“ ¥Ó We leave the verification of this equality as an exercise. from the previous equations we have Thus in both cases we have must satisfy the partial averaging property.

and so we have In conclusion. and any -measurable random variable must be constant (nonrandom).   £ ¾y z Ÿ™ º ª ÿ 2.e. the “trivial -algebra”. then you know the value of 1. . à '” ƒbG™ ! v …Aj ¥Ó ” ¶$ A supermartingale tends to go down. a submartingale tends to go up. ó € ª¤ £ ¾z hÆ ˜e  ‘y U½ Ÿ™ º ª ÿ ‹˜S%SU%€ Wž » y ‡ ʇ ¹‡¸‡ Ä · ‹˜S%SU%€ Wž » y ‡ ʇ ¹‡¸‡ Ä · ‹ ʇ ¹‡¸‡ Ä ·ž » ˜S%SU%€ W•y ‡ ” ¶  '” q­tuG™ ! v …Aj ¥Ó “ ” ¶$  ” ˜… ¶ ó ‹  ‡ Ž ŀ ‘y  £ »™ »™ » •™  ÿ ¡ ¸Ì“˜Gw¢ Ç¿Ÿ Í  ¤Â Á ¸Ì“˜Gw¢ Ç¿Ÿ Ë  ¤Â Á ¸ €  ¤Â Á fi˜Gw¢ Ç¿Ÿ ” `™ ” q¶ ” ¶ Ä ­€ · ™ É É É h h h ” ¶ £ y .4 Martingales The ingredients are: A probability space A sequence of random variables . is that constant which satisfies the averaging property The right hand side is .3 (Example from the binomial model. Each . We say that the process is adapted to the filtration . . This is called a stochastic process. Therefore. Martingales tend to go neither up nor down. the second condition above is replaced by . Example 2. is a supermartingale.e. by definition. This -algebra contains no information. with the property that . we set .) For we already showed that For . Conditions for a martingale: is -measurable.  ” ™i… ” ™ Ã Æ k¤ ½™ ª z š š 黕™ kG‚¢ À¿Ÿ “  ƒy U½„• Ÿ™ º ª  ¤Â Á € » ¾z ¼» ¹‡¸ SU@€ · à £ ™ ˜G‚¢ À¿Ÿ €ˆ  £ ¾y z Ÿ™ º ª  ¤ Á £”™  ˜GȐ¢ Ç¿Ÿ € ½™ º ª z ¤Â Á  ¶ r ¶$ ” fƒ' ” ™ ! v ” Aj ¥Ó  »WASs‚isl!TbD£¶ ¶ ‚ ‚A ¶A ¡ »p")‚is‚ss!`wAbG™ ™A ‚ A ™ D ' ¥A ™ ¦ B`"A „ $ . For each . If If If then then then is a martingale. i. If you know the information in . i. Such a sequence of -algebras is called a filtration. q » ™ q‚‚‚ q ! ™ q D wG˜µis¦`˜xG™ A sequence of -algebras .58 2. is a submartingale.

Cash outlay is $100.1 Binomial Pricing Return to the binomial pricing model Please see: Example 3. the interest rate quoted is per unit time. 7(1979). and Cox and Rubinstein (1985).1 (Pricing a Call Option) Suppose (interest rate). Prentice-Hall. The value of the call at time 1 is Suppose the option sells for $20 at time 0. .Chapter 3 Arbitrage Pricing 3. Sell 3 options for $20 each. 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. Borrow $40. Ross and Rubinstein. Cash outlay is 2. the strike price is $50). J. 229–263. Let us construct a portfolio: 1. and the stock prices are indexed by the same time periods). Cash outlay is 59 à Ð È›Ù Û ß Ð ›Ù Û à Ù Ð VÛ ß Ù hÐ VÛ if if if if Ô Ð Ø Ä C¦”× Ã Ù Ø Ã–Ã Ò ½× Ò ”× Ö ¥Ô ¹ Ð WÒ ÃÔ Ä Ð S%¹ ÑÏ Õ ÓÒ Ð ÄÔ Ä CÐ ¼ Ý Ä Ô i¡ÜÛÚ ½ Ú× Ð ›âÛÚ Ù á Þ ãÝ Ù Ý Ã Ä ¥§ã åä Ô ¹ CÐ ›ÜÛGÙ × Þ Ý Ú ÄÄ U¸ Ã Ä §Qã æä Î Î Cox.e. Financial Economics. 3. Options Markets. Buy 2 shares of stock for $50 each. . (In this and all examples.

i. 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. ( in the money market. at risk-free interest rate .. is also to be determined later) might be è Important Observation: The APT value of the option does not depend on the probabilities of and . Agent is a “small investor”. Unlimited borrowing. his/her trading does not move the market. No transaction costs. ë íì ë “ê Ä Ð ¥¹ VØ á ä Ä ‚“‚Èã ã ã ã ã Ä Ô ää Ä äÔ Qã Ä ¡Ù Û à Ð ä Ä ÈwÈwã ã ã ã ã ÄÔä Ä Ä ç¹ ä UÔ¥%äQã Ä h¡Ù Û ß Ð ë ˆê . é Î Î Î Î Î Î Î Î Î . 3. For this portfolio. ( is to be determined later). regardless of whether the stock goes up or down. îi„î‰ïuð¦î½ê ñ î ‰ï Buy shares of stock at time 0.2 General one-step APT Suppose a derivative security pays off the amount at time 1. (This measurability condition is important. where is an -measurable random variable. ( î‰ï î µê î ½ê Sell the security for at time 0. ë “ê ö ë ó î ï îê We want to choose and so that ûGGîibú˜@›°uV’A©¡V½b˜QSø ú ñ ò ÷ ùø ð ë ñø î ï ÷ î êú ò ÷ ù ú î ñ î ï ð î êø ú ò ÷ ùø ÷ ë ñ î Ui„‰yQ½AU˜@›°­¡’U¡ï ö ôö õë ó Then wealth at time 1 is î ñî ï ð î iG¡C‘½ê ò Invest negative). this is why it does not make sense to use some stock unrelated to the derivative security in valuing it.e.60 This portfolio thus requires no initial investment. at least in the straightforward method described below).

3 Risk-Neutral Probability Measure ! % $# " Let be the set of possible outcomes from by the formula ! coin tosses. Since .3) 3. Equa- ý ˆñ þ ý iê ú èø ë ú î ñú ò ÷ ùø ð ú èø ë ñø î ï ÷ î êú ò ÷ ù ÈAk“ê ö Gib˜QSf’ÈAk’x3¡@¡½b˜fVSø ú ø ë ú î ñú ò ÷ ùø ð ú ø ë ñø î ï ÷ î êú ò ÷ ù ƒüék“ê ö Gib˜QSf§ƒüéU’A©¡V½b˜QSø ú y’ÈAø ë ñ ðú è û ƒüéÿéø§ë“y’ÈAU“êñ ö ¡ï î ú ƒø ë ê ð ú èø ë ý iê þ î êúò ÷ ù ö ½b˜QSø ö ö ö  ¨  (2. Construct a probability measure B E 8 7 5 4 2 0 B @ 8 7 5 4 2 0  FA9) 6§3D  CA9) 6§31 % $# G #$ is called the risk-neutral probability measure. we have and .. probabilities. is known for a given . 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 . Thus.2) ý ˆñ  ¨  % $# . Define   ¢ ¢ ð ð Ñ û ø ë ò ð ù íð  ÷ ú è ø ë Gú ü駓ê ¢ @…  iÈA§“ê ¢ ð f  Qù ò ÷ ¢ ¦§˜òðùð¡¡ø¥búƒüU“y’ÈA„“Au’ÈA„“bú ú   ú é ø ë ê ð ú è ø ë ê ø ð ú è ø ë ê ¢ ¡¥ø ¤¢ ù í  ð  £ ð ñ ú ð ù ð   ú ø î ë bê ú ¢ ð ðÈ ø ë ð ú è î ñb˜òh…¡ø ƒÿé§i“y’¡¡Aøè  U“ê ¡ÈAø ë ê ú ú î ñú ò ÷ ùø ð ú èø ë ñø î ï ð ú èø ë Gib˜Q°f¡ÈAU’A©‰¡ÈAk“ê   þ ù   © 𠆠 û ¢ —…  ôö ò ð ù ð ò ÷ û¦ W§ú üéø ë ê   iÈAø ë ê   £ @ù Vù ö î ê ÷ú è  ¢ ÷ û Pë ê òfù cù I þ H ôö ú ( û û & ë ø '©)þ  UUû  'þ  —þÑ% $# G $# ù ö  ÷   𠅠 ð Qù ôö   ò ÷ ö îê  ¨ ¢ ¨ ©  î ¡ï Plug the formula (2.3) for into (2. i. We will return to this later.1) (2. We denote by tion 2. when Subtracting the second equation above from the first gives (2. is known (and therefore non-random) at that as well.4 says the expectation under  Then and .CHAPTER 3.1): (otherwise there would are like (2.e. We now also assume be an arbitrage opportunity).4) on . Thus the price of the call at time 0 is given by    ð ù ö   ò ÷ @Qù We have already assumed .

12 Under is a martingale. ëW ö “6ý ó ï UUû  íï  ¡Aø ö ï ûû ë î ï ûý úò ÷ ù %ˆñ ý R ˜fVSø ö ð ýˆñ ˜V¢ SUú…  …¥ø e“dý cR ˜fVSø ö ú ò ÷ ùø ¢ ð   Xë W V ú ò ÷ ù ý “ñ ò ð ð¡  ¢ ð…  gƒ½f¡  e“dý cR ˜fVSø ö ÷   ð ò   ÷ Xë W V ú ò ÷ ù ¢ ¢ ¢ ¢ ð…  ý ú ¢ ù ðí  ˆñ I ˜òhð`@ð¡ ¡ø ÷ ú @÷›ù°¡  e“dý cR ˜fVSø ö ð¢ ò ø X ë W V ú ò ÷ ù ¢ ¢ ý ˆñbú ¢   Š   H ø e“dý cR ˜fVSø ö ÷ Xë W V ú ò ÷ ù ý ` ë Wý Xë W V úò ÷ ù ¡ì a“6ˆñ Yˆ6ý FR kVSø £ ¦ b î 8 ý( U…ì  ˆñ ý Sk¡%S”Q T ý ý Rúò ÷ ùø The portfolio process is 3. ó ó î 8 ý( T ý ì  ý ý Sk@VS¨Q R ú ò ÷ ùø ó % #$ ó ö û ñ ò ÷ ùø ë W ñ ï Gú ý bú˜fVSCð “dý xø ý ÷ ý ˜QSø úò ÷ ù úý ñý ï ð øúò ÷ ùø ÷ ë Wý ñý ¥ˆU…h'ý U˜Q°­›ó“d“U…ï ó î ó . . the discounted self-financing portfolio process value -measurable.1 Portfolio Process Theorem 3.3. (3.3.û sý úò ÷ ùø ð ë Wý Xë W V úò ÷ ùøq ý ï ÷ Šbˆñ ý R ˜fVSuQ“6ˆñ Yˆ6ý FR kcS§r…@Qý ý R kcSø ö “6ý e“dý c6˜fVSø úò ÷ ù ë W Xë W V Rú ò ÷ ù Proof: We have Proof: Theorem 3.2 Self-financing Value of a Portfolio Process p ú °ë R ( h  G #$ % $# Each ý ý …ì †ï is Then each Define recursively Start with nonrandom initial wealth is -measurable. . (No insider trading).1) Î Î ý …ï Î Î .11 Under 62 is the number of shares of stock held between times . ý …ì ý Î is a martingale. the discounted stock price process . which need not be 0.2) (3. where and ù ÷ C¦i i 3.

e. is less than or equal to .11) (requirement (b) of conditional exp. then (4. the number of periods/coin-tosses in the model). satisfies the for each . the APT value at time of is for each w ê ûù ð ûû %v  G¥û  ù   ö i ý h„ ö 6…ì D“6y„ £ G #$ ¦ ý ` ë W ý T©v  Gû¥û  ù   û i … ý ý „ ö d†…ì  hDQ ûWSý‰ì ƒwAê w R ˜òQSø £ G $# ý ˜Q°ø ¦iê ¦ ` ú ÷ ù ú ò ÷ ù ôö ý ê i ì 9¦ ë ` ¦ ‰& R R w w hD ` ì` ì ë w w R û %& „£ w ‡„ £ R G #$ £ w ‡„ i  G #$ G #$ ý h„ ö 6…ì ƒ‡„ £ ¦ ý ` w ö ö ö ¦ & R w ì` w v  G #$ „£ ù ð ûû v  G¥û  ù   ö G #$ v  ûû GUû  ù   ö ù ð ²v ö i In this case. For we use the tower property to write ¦ & R w ð ²v ö w A ê Theorem 4.. the equation (4. for . Arbitrage Pricing Therefore.2) satisfies process x y is said to be hedgeable if there exists such that the self-financing value of . (Here. G $# 63 3.1 () A simple European derivative security with expiration time is an -measurable random variable .) (taking out what is known) (property (b)) ý ú Q©þø w ê ö ó `©þø ú ¦ 6 w ó ó v ûû ¥Uû  ù   ö w ó  ûû UUû  ë w A ê î ó  ó î i ó (4.1) ì v " ý ý R ˜QSø ö úò ÷ ù ý ñý ˆ¥…ï ý R ˜óú fcSŠð ò ÷ ùø ¦ Sý‰ì Dˆ6ˆñ ` ë W ý úò ÷ ù ý ï V cR ˜fVSø £ G $# …¦÷ X ë W e“dý úò ÷ ù ý ý R ˜QSø ö ¦ Sý…ì tˆ¥…ï ` ý ñ ý ú ò ÷ ù #$ ý R ˜@ó ›°ø £ G uð ¦ ý S‰ì Dëˆ6ˆ˜ý…ï ` W ý ñ ú ò ÷ ù #$ V cR ˜@›°ø £ G r÷ X ë W e“6ý ¦ Sý‰ì tý ` úò ÷ ù ý R ˜VSø £ G #$ ö ý `ë Xë W V ú ò ÷ ù …ì D“dóW ý e“6ý cR ˜VSø £ (Theorem 3. we call the APT value at time û ! € w‚þ ú –ë R w h  ê ï U¥û  ¡Aø ö ï ûû î ï Definition 3. w A w h Proof: We first observe that if martingale property i is a martingale. is hedgeable. then we also have ˆ ‚ i When .2 () A simple European derivative security a constant and a portfolio process given by (3.13 (Corollary to Theorem 3.CHAPTER 3.2) . i . i.12) If a simple European security .2) follows directly from the martingale property.4 Simple European Derivative Securities Definition 3.

e. (5. If the answer is “yes”. Set and define the self-financing value of the portfolio process by the recursive formula 3. is the APT value at time of .2) (5. Assume that (5.. Theorem 3. and set w R 6 û¦ý ` w ú ò ÷ ù Wxì ƒAê w R ˜@›°ø £ G #$ ý ˜fVSø ö ý úò ÷ ù ó ûSbý¡ì `ƒwAê w R6˜òf÷VSø £ G #$ ö 6…ì ƒw w 6˜QSø £ G $# ö ý ý 6˜QSø ¦ ú ù ¦ ý ` R ú ò ÷ ù R ú ò ÷ ù ó ó i úò ÷ ù ûû ˜VSø  ¥Uû  ë ë R ˜@¡Sø  î úò ÷ ù w w R be a simple European deriva- ý ó ó w 1 ê ö úò ÷ ù ˜VSø £ G #$ ý ˜QSø ö ¥›þ  ¥Uû  Q©þ©iê úò ÷ ù úý ûû ë øý w ó ó w ó  w û Uû¥û  ë ê ó ó  î ó w A ê i (5. i.5 The Binomial Model is Complete Can a simple European derivative security always be hedged? It depends on the model.1) and (5.1) (5. the model is called incomplete. let tive security. for each fixed . If the simple European security is hedgeable. the model is said to be complete.3) . 3. By definition.2).2).3) holds by definition of some value of . the self-financing value of the portfolio process .12 says that is a martingale. w A Starting with initial wealth is the process Proof: Let and be defined by (5. .3) holds for  úý ûû ë ø¦ý ` w UVþ  U¥û  Q—þ°6…ì ƒAê ê Theorem 5. If the answer is “no”.64 We can continue by induction to obtain (4. we have ûúý ûû ë øý úý ûû ë ø §U›þ  UUû  Q©þ%iê ö U›þ  UUû  Q©þ•ý úý ûû ë U›þ  UUû  Q©þø ó i î  ö i ó •©v  GUû  ù  D“’i ‘ ý ê ö ý ûT ûû Q € x ó ûúý ñý ï ð øúò ÷ ùø ÷ ë Wý ñý GU“k†'ý U˜fVSr¡“dˆ¥…ï ö “6ý ëW ó ë R hï  ¥Uû  ó ‰ï w ûû î ûû ûû îê ö î ë R w ï  U¥û  î ï ë R w ê  UUû  î ê ó w Aê  ¥Uû  “ê  ½ê ûû ë î ë R hï  U¥û  íï  ¡ï w ûû ë î R ˜fVSø £ G $# ¦ w ê úò ÷ ù ö îê w ý û û ë ø ë W ý ˆÈ ý UU ë ø “6ý ú ý û û ë ø ý ñ ðú ûû ëW û úƒéé  Vþþ  ¥Uûû  Q—þ©þ3“d“ˆÈèè  ý›þþ  UUûû  ëc©þ©þø3“dýˆêñ ö UVþ  U¥û  Q—þ©…ï ú ý U¥ ë ø “6ý ê ð ú ûû ûû ëW ëW . Therefore.14 The binomial model is complete.2: We need to show that We proceed by induction. and so for each . In particular. then there is a portfolio process whose selffinancing value process satisfies . For .

Arbitrage Pricing is a martingale under % $# We compute . We prove the first equality. the second can be shown similarly.û ú èø ë Wý GÈAk“diê ö úƒüék“diê iÈAk“6ý ê  ÷  bƒÿékˆ6iy’ÈAk“diüêø ö ø ë Wý ÷ ú èø ë W úú ø ë Wý ê ð ú èø ë Wý  ð  úƒéükë“dýiê iÈèAk“Wdiê ÷ I †  ø W ÷ú øë ý   ò ð ù ð úú ø ë Wý ê ð ú èø ë Wý y¢ —…  H bƒÿékˆ6iy’ÈAk“diüêø ö ú øë Wý ƒÿékˆ6iê   ÷ ÈAk“6iê  ÷ ú èø ë Wý ý ñ ˆñ    ð ý úý úò ÷ ù ý ñ ¥ˆñb˜QSøf𠈽• ø ú üék““i’Ƚk“diê ö ø ë dý ê ¢ ð ú Aø ë W ý W è ú W ƒÿéø ëˆ6ý ê ÷ ÈAø “6ý ê  ÷ ú è ëW  ñ ð ú è ø ë W ý úGˆbú˜@›°øu’ÈAkë“dˆxø ƒÿéüé3“d“¡úÈAk“6iêñ ö ý ñ ò ÷ ù ð ú è ø W ý ñ úúƒ økë“WW6ýiêð¡ÈèAøkë“W6ýˆ øë ý ýib˜fVùSrˆGýˆb˜@›ù°u’úÈAk“Wdˆüñø'…ï ö êú ò ÷ ø ÷ ú ñú ò ÷ ø ð èø ë ý ý ú ý ñ ý ð ý U˜@›ù°¯iúÈAø “Wdý ñ ý ï ö ï øúò ÷ ø ÷ è ë ú èø ë W ÈAk“dý ó û úú ø ë Wý ÷ ú èø ë Wý ò ÷ i5ƒüék“diê   iÈAk“diê   ø ù Qù ö ý ê ó Since will be fixed for the rest of the proof. For example. 65 . we simplify notation by suppressing these symbols. In particular. we write the last equation as úý ûû ë G›þ  U¥û  Q©þø ÷ û½búƒé  ›þ  ¥ûUû  ëQþ©kë“dýiê   iúÈè  ýVþ  Uû¥û  ëQ—þkëˆ6ýiê   ø ò@ù Qù ö ú ý û ø W ÷ û ø W ú ûû ë ø ý ë W ú Uý›þ  U¥û  Q©þ°¦S…ì `D“dýiê ë R ˜ò÷VùSø £ G #$ ö ¥›þ  ¥Uû  Q©þ©iê úý ûû ë øý î 8 ý( aý ê ý 6˜QS¨Q T R ú ò ÷ ùø ¦ 6 ö ö ö ý iê ú ò ÷ ù ý R kcSø ¦ ý S‰ì ƒAê ` w R ú ò 6˜›°ø £ G #$ ÷ù w ý `¦ë Wý ` w Rúò ÷ ù ¡ì 9P“6…ì ”1ê w 6˜VSø £ G #$ £ G #$ ¦ S ý `ë Wý Xë W V Rú ò ÷ ù ‰ì Dˆ6iê e“6ý c6˜VSø £ G #$  ûGƒé  ›þ  UUû  Q©þU“diê ö ƒé  ›þ  U¥û  Q©þk“6ý ú ý ûû ë øë Wý ú ý ûû ë øë W ú ý ûû ë øë Wý ú ý ûû ë øë W Èè  ›þ  UUû  Q©þU“diê ö Èè  Vþ  ¥Uû  Q—þkˆ6ý ó We need to show that ó In other words. Note first that CHAPTER 3.

66 .

we have: The Markov Property Chapter 4 67 ß † Úß × Ú ß oo † Úß aá Ð Ý ß –Ù Ù Ý Úß ½× Ý ß –Ù á Ð Ø Ú „ „ .1): — — – ˜ †f 0Ô Ð ™d j i ãVç Ð 6Ò 0ԗ†–mÐ kkhlfÙ Ð d™ Ò æ ™n j i‰g e ú ý ƒé  ›þ  i  ú ý ƒé  ›þ  ûû GUû  ù   v  Ð Ø × Ò Ô ¹ ˜Ð WÒ 0ԆfÐ S%¹ …Ï — – Õ — – ÓÒ Ð ûU¥û  Q©þk“6ˆy“Èè  Vþ  ¥Uû  Q—þ3“d“ñ U›þ  U¥û  Q©þ•…ï û ë øë Wý ñ ð ú ý ûû ë øë Wý ûUUû  Q©þU“diy“Èè  Vþ  ¥Uû  Q—þkˆ6iê ö ú ý û û ë ø ý û ë øë Wý ê ð ú ý ûû ë øë Wý ¦ ý ` w R ú ò ÷ ù úò ÷ ù ý ö i  6…ì h1ê w SkVSø £ G $# ý ˜QSø ö iê w ê Notice that Ð Ý V Úà †á Ò xÐ Ý Q Úà †á ”xÐ Ý ¡ Úß †á Ò ¥ç Ð Ý … Úß †á à o – ß o yÐ w à o ç ß o Using these values. Working backward in time. 4.1 Binomial Model Pricing and Hedging Example 4. and the value and portfolio processes are given by: 4.Ý à¡ Úß × ½Ý ã Ò ‡vå —˜–zÐ Ý à¡ Úß o ½Ý ã o †á Ù Ð ÝÝ à Úà „½× ãã ÚÙá Ò w …— – v††x — Ô Ò – å †hÐ w  —  ƒu ˜– Ô Ô — æ — –  — — Ô  — ¹ ¹ Ð  ƒC0Ԙ– { – å ˜xuÔ ˜– } æ Ð Ø á Ô – –  — –  — Ò xÐ  ‚u0Ԇ– { ‚u0Ԇ– } æ Ð Ý Úà Ù á { } à ß ç  — Uç u0Ԇ– } Ôæ Ð €Ý ¡ Úß o á ™ n ~Ý … Úß o á ™ d cÕ ç |ç Ð Ý Úß Ù á { { | — – ˜z — e Ý Ô u Ú× ã t u t vt‰ s r p o bhq Ø Ð o †á û%ùfðv  ûG¥û  ù   ö û ûù %@ð Recall that is the given simple European derivative security.1 (Lookback Option) Consider a simple European derivative security with expiration 2. with payoff given by (See Fig. we can now compute: The payoff is thus “path dependent”.

2 (European Call) Let with expiration time 2 and payoff function „ Note that — ŽÝ h•GÙ ‹  { Ý Ú Ô Ý Ú ¹  Œ Ú Ùo ~©Œ ¹ GÙ ‹ Ùo } æ Ð ©ŽŒGØ ‹ Ô Ý Ù  { Ý Ò ŽÝ ¹ h•ŒÚ o ‹ Ùo ~©Œ ¹ Ú o ‹ Ùo } æ Ð ©ŽŒÚ ¥‹ Ý Ô ã ŽŒÚ Ð ©ŽŒÚ o ‹ Ý e Ð u  Œ h× Ô — U‡ — ç – å Ð  ƒ Ùo { – æ — æ  Ù o } æ Ð Ø á Ô – xÐ  – — Ùo { †–— Ù o } æ Ð Ý Úà Ù á Ô – æ — æ Ð  †— – { ç Ùo cUç — Ùo } æ Ð Ý Úß Ù á – à – ß à ß ç Ò xÐ Ý V Úà o á Ò gÐ Ý Q Úà o á Ð Ý ¡ Úß o á Ò Uç Ð Ý … Úß o á Define etc. Ý ••ŒÚ u 6‹ Working forward in time. Then S2 (TT) = 1 S2 (TH) = 4 S2 (HT) = 4 S2 (HH) = 16 .1: Stock price underlying the lookback option.Ùo Ð ™n — Ý Ô ã Ú× Ð aá e o o Ð vd™ Ò æ Ð Ø × Ò ÙŠ Ð SÒ Ùo Ð W"¹ ŠÏ Õ ÓÒ Ð ß Ú ç Ú ç Ý Ú Ú ã Ú Ú { çÚ { ß Ú Ú Ò Uç Ð Ý ß…ß GÙ á ‰ ç †–— Uç Ð eÝ ß „Ù × Ý ß Ù ½Ý ß GÙ ˆ bÝ Õ |e1Ý …ß „Ù × Ý ß Ù Ð Ý …ß „Ù – — – Ú Ý ã Ú { çÚ { Ú Ú Ò gÐ Ý à „Ù á ‰ ç – ˜zÐ –Ø × Ø „ ’Ø ˆ 5Ý Õ |eAÝ à GÙ × Ø Ð Ý à GÙ ˆ „ — Ô Ú ‰… —Ô Ý ã Ú { çÚ { Ú Ú Ò – å †hÐ Ý ß „Ù á †v0ԆhÐ GØ × Ø „ §Ø ˆ bÝ Õ |h~Ý ß GÙ × Ø „ Ð Ý ß GÙ ˆ ˆ „ „ à Ý V Úà × ½Ý Q Úà × ã ß —˜–zÐ Ú Ý àV Úà o á ½Ý Q Úà oo á Ð Ý à –Ù ã ß o Example 4. we can check that 68 to be the value of the call at time S =4 0 Figure 4. and consider a European call . S (H) = 8 1 S1 (T) = 2 when .

2 has this. then in period 66 there would be 67 possible stock price values (since the final price depends only on the number of up-ticks of the stock price – i. " ( A ú é ø& ú è ø& é èø& ú è èø & ƒ—üéGFê  ȃüéGcê  ú wx„cê  ȓxkcê 3. the option in Example 4.. Then ö " . a three-month option has 66 trading days. More specifically. Example 4. There are three possible ways to deal with this problem: 1.e. equations of the form " has elements. In period 2. We’ll get to that. We have. Simulation.2 Computational Issues ˆ i ˆ ! For a model with periods (coin tosses). We could repeat this several times and take the average value of as an value of approximation to . Approximate a many-period model by a continuous-time model. we could simulate coin tosses under the risk-neutral probability measure. we must solve For example. 2. ™ f  ù ˜ — ë hh“• ” P” ˆ ý ê û¦ ý ûû ë øë Wý W§ú é  Vþ  U¥û  Q—þkˆ6iê   ÷ Èè  ›þ  UUû  Q©þU“diê   £ ù Qù ö U›þ  UUû  Q©þ•ý ê ú ý ûû ë øë Wý ò ÷ úý ûû ë ø Œ Ð u× —  Œ 㠌 – Ð ç Ò ‡¦ Ò Ý h•ŒG½¹ St‹’½©Œ ¹ G½Su ‹ Ð ©ŽŒ©u ‘ Ý Ú ¹  Ú Ù e u ã Ý ¹Ú Ù e  ê {G $# ( R kcSø ö ½ê úò ÷ ù î ( úùø& ú ø& ùø & US¥D›  ©Žœ¥D›  ú ’ °%D› îê ( ê AŠG $# ú ( `©þø Aê ú( ûû ë ©šþ  U¥û  Q©þø ö þ Ý Ú ••Œ©u ‘ Let be the number of shares in the hedging portfolio at time when . .CHAPTER 4. rather than . for example. Look for Markov structure. that (  1 ™ f  ù ˜ — • ” ” ë vyh“–PUˆ ’ ’ †“ and so we could compute by simulation. — U‡ — ç – Ð Ø å Ð  ƒ Ùo { – æ — æ  Ùo } Ôæ VG‹ Ô – Ú Ò zÐ  †–— Ùo { †–— Ùo } æ Ð Ý ¹ –Ù ‹ Ô Ý Ù { ç Ò – æ — æ Ð  †–— Ùo F¥ç — Ùo } æ Ð UŽÚ ¥‹ – Ýç – Ý ç Ýåç Ò zÐ –Ú o ‹ Ò zÐ ˜8æÚ o ‹ Ò Uç Ð UeÚ o ‹ 69 4. Then we can use calculus and partial differential equations. For period . We could store the . rather than four possible values If there were 66 periods.2 has three possible values . If each day is taken to be one period. then and . heads – so far) and hence only 67 possible option values. The Markov Property In particular.

the distribution of conditioned on . we have ëW “dý ý (The Markov Property). This process is said to be Markov if: ¡ f£¡ h $ ¢ $ e Ñ  ó î 8 ý( a…UQ T ý ì Ÿ ž ú ¤ ‰ ì  ¡ø ! ú ó ¤ P úë W ø ` –ˆ6ý ’†¨’G $ Ÿ ¢ xCŸ ì  ¡ø ! ¨ ó   ö ý Ñì Î Î î 8 ý( aý T ë R c  ó . For every (Borel-measurable) function for which . ó a‰¥Q T ý ì ëW “dý ùfðí"  ûG¥û  ù   ö û ó ôö ± b« ª ÷ °¼»º ck˜bY` ö ® ¨ ú °ø ö µ¡’¨ ó ó ó ± b« ª µ ª ² « ´³G $ ª ¬ « Ž­G $ ú °ø ö ½¡¡¨ G$ T U ó © hv ` ý Q The stochastic process is adapted to the filtration ..e.. i. then it holds for every .3. then the equations in (b) and (c) are the same. i. (c) implies (b). Let be a filtration under be a stochastic process on . (Since we don’t need to assume that  ¶ b ý ` ˆ6ý ëW ó $ € ó ¡ ‚’  (d) (Agreement of characteristic functions) For every ¶ ý ö v…ì ` “dý ëW .) For every for which . and conditioned ì Definition 4. Let Q . we have ëW “dý û¦ ž € W S‰ý `h‚¦ë“dý £ % $ ¦ ‰ý ` ú W ø § ó –ë“6ý Q6$ ó £ G $ ¦ bý¡ì –“W6ý ø § ó ` ú ë ó Q6$ £ G $ ó ¦¡ ¡¦Ÿ ú $ø ó W‰ý v°“6ý ¡¨ £ û¦ `úë W ø ó ó  G$ úý ` ž € ë W ø ö U…ì h‚¦“6ý …% $ ö ö ¦ ý ` ú ë W ø ö Sxì –“6ý ’¨ £ ® ® ¯ ý ö S…ì 1“6ý ® ë W Ÿ € ¥“ž (a) (Agreement of distributions). and in (c) we assume this condition only for functions of the form . we have ¨ ©  « ª hG $ û ¯ ®®® ë W ƒ6ý Aˆ6ý ó ® ¨ ó µ ª ² « b´·G $ ª ¬ « •­G $ ¡ ‚©  $ € (c) (Agreement of Laplace transforms. 4. For every .3 Markov Processes Technical condition always present: We consider only functions on I and subsets of I which are R R Borel-measurable. However in (b) we have a condition which holds for every function .e.1 Different ways to write the Markov property (If we fix and define . we have ¡ m£¡ † $ ¢ $ e w¨ (b) (Agreement of expectations of all functions).70 4. A main result in the theory of Laplace transforms is that if the equation holds for every of this special form. we only consider subsets of I that are in and functions R such that is a function .) © . For each on is the same as the distribution of i .1 () Let be a probability space.) ~` b« µ †’G $ ª` ± ù ` ° ¾ ½ ¼ 3¦´P ` b« µ †` ª ± ù ¦ð ¹ “ ¸ where .

All these Markov properties have analogues for vector-valued processes. we could just as Remark 4. We write these apparently stronger but actually equivalent conditions below. ¡ Á $ 4 d W W 6ý ûû ë W ž € ë W £  U¥û  “dý ‚¦“6ý ý ‰ó  ¥Uû  ¡ ¦“dý ž ó ûû $ Á ë W À ž   (B) For every function for which . The Markov Property G$ 71 4   i %$ i ú ý ¡  ø ó   . û¦ W‰ý ó ` ó Å YÄ X 7 Ç 7 YÄ « Å Æ Å hYÄ W ††|W ÈÈÈ Ç hYÄ « V µ ª £ Æ Å ûû U¥û ú `ûûûø ¥ý d¥USfG $ G$ ó ¦ ý Å ö S‰ì ` X 7 eÄ 4 ¡ ‚u 4 d    $ € W Ç 7 YÄ « Å Æ Å hYÄ Ç hYÄ « V µ ª £ Æ Å W ††|W ÈÈÈ G$   (D) For every we have © mh` û¦ WSý ó 7 YÊÇ 7 Y« Å Ä Å Ä ` Å 7 eÄ Ç 7 YÄ « Å Æ Å hYÄ Ç hYÄ « ª £ Æ Å W ††|W ÈÈÈ Æ Å Ä Ç Æ Å Ä ª` hY)fhY« †’G $ G$ ¦ ý ö S‰ì ` Å Å Æ Å 7 YÄ Ç 7 YÄ « W ††ÉW hYÄ ÈÈÈ 4 ¡ ‚u 4 d    $ € W ú ½ U¥û  “6µ¡ø ö ý ûû ë Wý   Ç hYÄ « ª £ Æ Å G$   (C) For every for which û¦ WSý ûS‰ý ú 4 6ý  ¥Uû  “6ý ¡¨ £ G $ ö Sxì ú 4 6ý  U¥û  “dý ¡¨ £ G $ ¦ ` W ûû ë W ø ¦ ý ` W ûû ë W ø ó ©hó `ú 4 6ý  UUóû û  “6ý ’†¨’G $ ó ¡ m¢ ó¡ t‚¨ W û ëW ø ` $ $ e 4 ó ó ûû ë W ¥Uû  “6ý ø £ % $ ö S‰ì ƒ‚–ú 4 6ý  UUû  “6ý ø £ ¦ ý ` ž € W ûû ë W ó ` ž € h‚– 4 6 W ú ý W ††ÉW ÈÈÈ ó  ó ú ½ U¥û  “6µ¡ø ö ý ûû ë Wý   ó 4 ó ¡ ‚¿ž $ € (A’) For every . 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. . Let be a positive integer. we have ¦ S ý d¥Uû £ ` û û ó CHAPTER 4. we have .appears. Consequences of the Markov property. Conditions (a)-(d) are equivalent. form (a) of the Markov property can be restated as: Ÿ € ’¿ž where is a function that depends on the set . where the û¦ W6ý ó ` 4 6 W ý ž ƒ©€ 4 6 W ý ó  ûû ë Wý ž € ë W U¥û  “6ƒ‚È6ý ó (A) For every %$ . we have Remark. £ %$  ú ø úý ` ž € ë W ø Gý “  ö ¥…ì ƒ©¦“dý …% $ ó ¦ ý W ý ž ö S‰ì ` 4 6ƒ‚€ ù ÷ r¡i ó For every . every expression of the form can also be written as function depends on the random variable represented by in this expression. For example.1 In every case of the Markov properties where well write for some function . ú ø Gý ¡  Once again.

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 — Ý YÔ × Ô { Õ YÔ × { Š Ô Ú× çw Ð Yaá Ô Ô × is a function of . they are equal to each other.1) × %$ u Ï .2) . form (a). If we want to estimate the distribution of based on the information in . Equation (3. conditioned on use the tower property on the left.2) is the case of .1) (Markov property. form (b). and (3.3 It is intuitively clear that the stock price process in the binomial model is a Markov process. Note however that form (b) of the Markov property is stronger then (3. and this is property (A) with ú ø Gý QÍ Now take conditional expectation on both sides of the above equation. Then (a) also holds (take ö ö ö ö ö ö ö ö %$ ë “ý u h ö ó £ ). ˆ À are equal to the RHS of (3. (3. is a function of . We will formally prove this later. to obtain ö ó û¦ `úë W ø ûúë W ë W S‰ý W“6ý ¡  G–“6ý ø “dý ó ó ó ó ¦ ý 6…ì Sˆ6ý ` ú ë W ø ûúë W ë W ¡  G–“6ý ø “dý ž $ £ G $ ó ó & W ¦ ý ¡ì 9Y“dý ` ¦ ë W ` ú & W F6ý ø cdý ž $ £ G Ì$G–“6ý ø “dý ž $ £ G $ ûúë W ë W ó ó cdý ž ó ¦ ý ì 9¦ “6ý ì ú F6ý ` ë W ` & W ø & W $ £ G Ì$Gú “6ý ø “dý ž $ £ G $ û ëW ëW ó c6ƒž ë óW ¦ bý¡ì `9‰ëˆ6ýxì úGcdý ¦ W ` & W ø & W ý ‰–ˆ6ý ø “dý ž $ £ G $ £ G $ $ ú ë W ó c6ý ž ó “dý ž ¦ ý ì `vú F6ý ø & W Pú “6ý ø ë W $ £ G $ $ ë W & W ý `&W ž € &W ëW ž € ëW ‰ì tócdý ‚Ãcdý  ó “dý ‚¦“6ý (Definition of conditional probability) (Tower property) (Taking out what is known) (Markov property.) ó Ùe ½bu × u ¦ 6 ¦  × Ý Õ |eÚ Ð u × Ý Ó ™ n { ç ý ó ý `&W xì hF6ý ó ` & W †F6 ¦ ‰ ý ¦ S ý Sˆ6ý ¡  G–“6ý ø “dý ` ú ë W ø ûúë W ë W ž € & W ‚ÃF6 ž € & W ‚ÃF6 ž $ £  u G$ u Ï Ò 6 × Œ Ð ••Ú Ó ÝŒ Ù eu Ý ½bh Ú× Ó } º Ñ Ð ý ý ó ¦ ö Sý ó {   ëW ˆ6ý ëW ˆ6ý Ï ™d Ú Ð ó ` & W †cd ž € ‚¦ d W ž € ‚Î 6 W ý  bu ó ž € & W ‚Ãcd ë “ý ë “ý Ï Ò  bu e Ù ½ × } ºÑ Ð Ó ó ó ý ž $ £ £ £ ó %$ %$  G$ ëW “dý ó £ ž € ‚£ d W Proof that (b) Consider (A).1)). For example.) (Remark 4.72 § 6$ ¨ ˆ À Ë c Since both and Example 4. (with in (A)) Assume (b). . the only relevant piece of information is the value of .2).

Recall that by the first toss. × 4. contains the four sets Þ — é c F è Ò ä à ß à ã âÐ å ä à ß ß ã âÐ à i Ò –Và Ò Q–çæ)ß Ò –¡ß Ò …Gfáß Ù Example 4. i. the -algebra generated Ú € ’­Ü Definition 4. (We are using form (B) of the Markov property here). if we take and from and ß ß ßã Ð Qà Ò …„ìCë Ï d o ä–¡ß à Ò ß…GmÐ ß ßã Ò ä à à ß ä ß ß ß ½ Ò –Và Ò ¡Gã Ò Qà Ò …G㠗 é è e e ê Þ ú Ù ¤ ‰ Ï ê ì  ¡ø ! ê Let be the -algebra of sets determined by the second toss. but o d or ). and let algebras of . which we can determine with a bit of work. Assume be random variables on a probability space Ý Ú× o Þ Note that and Þ ä  ß ßã Ð …„²ë — ñ †d Ð Ýn Ù Ï are not independent (unless .4 Consider the two-period binomial model. For example. We say that and are independent if for every and ó Ù € ‚Ûž Ø ÃÏ Ò Yaá Õ Ô Ô û ú Üø $ žø GŠ¥…% Pú ¡…% $ ö Š‚f¥…% $ ú Ü Ý žø Ò  Ø Õ × Ò YÔ á } º Ñ µi Ý Õ |eÚ Ù {ç Ô Ý Õ |eÚ } Ð Ôº Ñ Ø Ý Õ |eÚ { ç { ç Ð Õ i Ô eÔ ú ¤ P ì  ¥ø ! Ð Ø aá Õ Ð Ú ó Ù Ù Ï ØÕ ì Ö and be sub. In other words.4 Showing that a process is Markov Í Ú Ù The following lemma will be very useful in showing that a process is Markov: ò Lemma 4. Ù is independent of a -algebra if . we have ó .e. Let be a sub... i.. one of the sets in from and from d is .e. The Markov Property The value of this security at time 50 is  9 73 because the stock price process is Markov.CHAPTER 4. contains the four sets is the -algebra of sets determined Í ú VÍ ø Ù Ï Í We say that a random variable by .-algebra of . then are independent.2 (Independence) Let be a probability space. the -measurable random variable can be written as — — — ÝeÝ Ø Õ Û Ò 6—d— Ò Ù ÛâÚ Ø Õ Ú× ×ØÐ Ý Ø Õ Û Ò 66— Ò Ù ÜÛÚ Ø Õ á Ø †á Õ for some function .15 (Independence Lemma) Let and . For example. then ä  Ý o Ú× Ù — o d Ð Ýn o { d Ú { o d ä à –¡ß ß ßã Ò ç †Ð „ïÐ –mÐ d d Ð Ý td { d bÝ hd Ú n n Ú †d Ð d Ý td ñ n o o { ó Ú Ð Ý ë Ú íÑ Ý ß Ú { ß o ì Ú Ð Ý ë Ú íÑ Ý ß Ú íÑ d Ð Ý … Úß Ð Ý ß íÑ Ð Ý ß… Úß í Ñ Ð Ý ë “ß Ú í Ñ d äð߅„ãïЊä Ø × rÐ ¡âÚ o ß Ï Ý Û î S‰ Û ã × × o Ù Ï o Í íÑ ë o î æß Ú Ù íÑ ê Then and from . is independent of . If we take .

u Ý h Ú× This shows the stock price is Markov. Then G$ Remark. Since is -measurable. of .is -measurable. and define Ù Ù ó ò 74 Î Î ºÑ is independent of . Define the running maximum of the stock price to be " Examples: ú( (ñ ©‡„  cAø ( v› " Consider a simple European derivative security with payoff at time ý 4ÿë ô ý û 4 ñ µaÿþ ƒ‰üŠh„ ý û ö  bu Ò 6 ÝÙ e °½bu Ú× × Ó}  bu Ò d ÝÙ e –½bu Ú×  6u Ï Ò d ÝÙ e b½Su Ú× Ó} Thus Ó} and are equal and form (b) of the Markov property is proved.5 (Showing the stock price process is Markov) Consider an -period binomial model. Indeed. — Ý UÓ Ú ú — Ý uh× u × { Ó Ú Ó n AÝ h© ÚÏ Ó d Ý øÚ ×  ÷ 6Ý ø Ò Ò ˆ Ú ù } ºÑ u h× } 6u Ò I u × — Ùe ½bu × H Ó º Ñ Ó Ï Ý ˜u × Ó Ú n Þ { 1 Ý ú ÚÏ Ó n Ó { ~ u d Ý × k • ÚÏ Ó d Ð 6u × d–½bu Ú×  Ò Ý Ù e Ð Ð Ð Ð  bu Ï Ò d ÝÙ e b½bu Ú× ˆ Ð Ó} ÷ ºÑ ºÑ ºÑ Ï ºÑ  time and define and .5 Application to Exotic Options Consider an -period binomial model. but we have also obtained a formula for as a function of . so that à‚Чٽbu Û Ó Ð e ˆ ö û Gú¡Žòø¡  ö cõٝ‘ò  ›ó £ ¦ ` ú ø ó ø ó VhG $ ôö ú ô “  ø û Gú ô d Ð Ý ú ˆ Ú Ó º Ñ Ð Ý YÒ ˆ Ú ù º Ñ âÐ Ý ú Ú × ú Ý •ŒÚ ÓzÐ Ý •ŒÚ ù ú ú eÒ â Ýç –È{  Ú ˆ u h× Ùe ½bu × âÐ ó ˆ Ï â ˆ ú ° ô d ¥ ó ø › Let be a function of two variables. capital letters denote random variables and lower case letters denote nonrandom variables. Then if and is independent of . Fix a uh×yÐ ø ÷ â ß Ð Ùe fQ½bu Û @Ð Ï u ˆ be any function and set . Let . Define . u 4. In this lemma and the following discussion. 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 Þ u Ý h Ú× — and Ó ÷ ø depends only on the st toss. This is a special case of Remark 4. Example 4.1. if . Then The Independence Lemma asserts that Ó} Not only have we shown that the stock price process is Markov.

although that does not matter).16 The two-dimensional process the risk-neutral measure I . We have (Knock-in Barrier option).. The Markov Property is independent of indicates the maximum of two quantities. (Here we are working under Lemma 5. „  úý „  ý „ ö ý ñ ¢ ý  ø ÷ ú ñ  ø ý ñ ¢ ¡¨   ibú ý µ¥ ý ûúú ø G5µ°  ¢ ô d ° ¢ „  ñ  ø ý µ¡’¨   ö ú ý  „  ø¡¨ ib½1¡ 6 A¡¡¨   ö ÷ úú °  ø ô °  ø úú ° dø ô ° dø ¨ bµÃA d ¦A’”G $# ôö ú û ú ý d  ý ý dø Gú5Uˆñ Aøyy„  ˆñ A’¨ ö ú Wý ñ „ë“6ˆfýh„  “6ˆA’¨ ö –“dh„  “6ˆA’¨ ë Wý ñø úë Wý ë Wý ñø ú 6 ¡’¨ ° ø ý …ì ô dø ú dø  ö ú ¢ ö Aí% $#    ö µ  ö AÑ% $#  ý ˆñ ëW ˆ6ý ñ ôö d ë Wý ñ  ý “6ˆfy„ ö “dh„ ë Wý  î 8 ý( 6Uh„  ˆA¨Q T ú ý ý ñø Stepping back one step. we have i úý ý ñ Gh„  “Aø „  ë W ñø ˆ6ý A’¨ £ G $# The Independence Lemma implies ø “ Define Proof: Fix . we have proved the Markov property (form (b)) for this two-dimensional process. so 75 .û¦ ©§ú ë R ( „  ë R ( ñ ¢ø ( ›  ÷ ˆú ë R ( „ ’  ë ½ ë µø ñ ñ R (    R (   ¡ ( › ( v› £  £ G $# ¦ Y ë R ( ƒ  ‡„  c ` ( ( ì © ú ñ xø ò@›ù ÷ ù ò@›ù ÷ ù ö ö ë R ( A ê û ú ( ( ñø ( ( §3‡„  cAˆ› ö Aê ûù ð ûû %í"  G¥û  ù   ö ý iê % $# i  S ¦ ý `ë Wý …ì “6iê £ is Markov. we have òQù iê ÷ ù ö ý ý ê ý R ˜fVSø úò ÷ ù ñø ` ë W ý A“  ö ¦ ý ì ú “6ý ° ô d ¥   úÈ ðŠcA¦©§¥£$ ö ©g„  cAi› Î ( ñ ø ¨ ¦ ¤ ¢ ú ( ( ñø ( W ú   ð ( „ ú ( ( ñø ( È¡¡g¡ø ö ©g„  cAi› Î W At the final time.. We have i where  and d CHAPTER 4. Since is a martingale under . be a function of two variables. Let . we can compute G #$ Continuing with the exotic option of the previous Lemma. Let (Lookback option). Since the RHS is a function of . P the second equality being a consequence of the fact that . Let denote the value of the derivative security at time . .

the hedging portfolio is given by the usual formula. Since this is a simple European option. which in this case is  t¯ ú ô d ° ¢ ú ý ý ñø ý ¥y„  ˆx©˜› i ø ë Wý ÷  ° °  ø ë Wý ò ÷ k“6˜›   “ú ô A   A¡3“6˜›   ¬ fù cù ö ú ûú §°ë R ( ‡„  ° ô d ¡ øý ©˜› ë R ( c ñø A3ë R ( › ö ë R ( A ê ° ô 6 ¡ ¦ ô § 6 ° ¢ ú ø ( v›  ÷  ° °  iú ô 1   A¡ø ( v›  £ òfVù ÷ ù ôö ú ø %ë R ( v› The general algorithm is so that This leads us to define 76 .ñ ð   ý bú †¡ø ú ý ý ø ë W ý › ð ú ý¢ „  ú ý ñ   ý ñ   ø ë W ý ý ¥y„  ˆñ ¢ 3“6kg§Uh‚Uˆ½¡ø  “µ¡%ˆ6˜› ö …ï and the value of the option at time is .

Again a function is specified. the holder of the derivative security can “exercise” and receive payment . and the wealth process i American algorithm. Thus. Define by backward recursion: Now consider an American option. In any period . i ú ý ñø ý U“A©˜› Then is the value of the option at time . . the hedging portfolio should create a wealth process which satisfies almost surely. i ûù ð ûû %@í"  GUû  ˆ  ù   ö ú ý ñø G“A“  i    ýñ ð   ˆbú ¡¡ø ú ý ñ ¢ ø “6ý zVú ý µ¡ø “6ý › ö ý ï ë W › ¢ð ñ   ë W  i Y x i ú ý ñø  G“A“  ý ú °ø ½¡“  þ ý Uƒû ó ú °ø ý ö µ¡•˜› ú °ø ( ö µ¡i› ú ý Aø ý › ñ Then is the value of the option at time . ú °ø úú ø ë Wý ÷ ú °  ø ë Wý ò ÷ µ¡¡   bµ° ¢ 3“6˜›   “µ1¥kˆ6˜›   ø fù Qù Þ 77 ú ý ñø GˆA¡  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.1 American Pricing Let us first review the European pricing formula in a Markov model. and the hedging portfolio is given by ÷ ûW§½° ¢ këˆ6ý˜›   ÷iúµ°A¡ø3ë“6ýk›   £ òù Qù ¦ ú ø W   W ú °ø ½¡“  ú ( ñø •cA¡  ö Aê ( ú °ø ý ö µ¡•˜› ú °ø ( ö µ¡i› " .Chapter 5 Stopping Times and American Options 5. Let be the payoff of a derivative security. Consider the Binomial model with periods.

1 See Fig. 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 . å w— ç Ð ä ç å †— ç †bhp Ò w ãsr Ð ! Ý ˜æÑã ÚÔ Ò #%Ý †–—†w Ú — { Ý — "Ô Þ s r Ý Ú Ùo t˜æ ˜– Ú — Ùo æ Qbhp Ð ˜8æ¥Ø ‹ – — v– ˜w Ð Ð Ð Ð Ð ¥•Ú ¥‹ Ý Ù Ð Ý ¹ GÙ ‹ Ú e ä Uw ã s r Ò †bhp Ô ! %¹ s r Ý ã ÚÔ Ò # æ — Ùo {|ç — Ùo æ Þ bhp e ¹ – æ ˜– — Ô sr – Ò Þ bhp ! "Ô¹ Þ s r Ùo æ Qbhp ݐ U’ã ÚÔ Ò $ç — Ùo # { – †— . Begin with initial wealth as follows: „ e å †w— ç Ð Ø Ø 78 Figure 5. Then Let us now construct the hedging portfolio for this option. 5.Ð Ð Ð Ð Ð Ð Ð Ð – — æ ˜– – — v– ˜w w æ ˜– 㠗 Ð Ø 'c– ‡ — hcØ w ã &Ð ç { Ý æ ã ç bØ Ñcå w — eÚ æ { Ô cØ „ ¹ „ Ý–Ø × Ø ›Ø bÝ |e|Ø Ý à „Ù × ã Ú { çÚ { Ú ˆ „ Õ Ý Ú Ú eÝ à „Ù × „Ù„ ‹ w æ „ —˜– ã Ð VØ 'c– ‡ — h{ Ø w &Ð ç„ Ý æ ã ç bØ Ñcå w — eÚ æ { Ô cØ „  „ Ý –Ø × Ø ã’Ø Ú5Ý { ç Ú { Ú Õ |eÉØ Ý ß „Ù „ × Ý Ú Ú eÝ ß „Ù × „Ù ‹ ˆ „ „ „ ˆ e ݌ ©’ã ÚÔ Ð ••ŒÚ “Ð ••ŒÚ o ‹ Ý × Ý Ð ¹ {ö Ò Ùo Ð ™n Ð Ø Ð vd™ Ò ÙŠ Ð SÒ Ùo Ð W"¹ ¦Ï Ò æ Q”× ÓÒ Ð Õ Example 5.1.1: Stock price and final value of an American put option with strike price 5.

equivalently. ûû G¥û  ù   ö i ó ýƒ(0˜òf÷VSøC’Uˆb˜Q°fQˆ6ˆA”…cý ˜QSø ú ù ð ú ý ñú ò ÷ ùø ð ë Wý ñø ý ï ÷ ú ò ÷ ù úý ñý ï ð ( ð øúò ÷ ùø ÷ ë Wý ñý ¥ˆ˜…ý )ý ¥kcS­¡ó“dˆ˜…ï ó ö ëW ö ˆ6ý ó – –— w ††x ç Ú &Ð { Ú ç Ò ã Ð Ý à –Ù 'c0ԗ ¹ ~Ý à –Ù 0ԗ ã Ð æ &Ð { Ò çã Ð Ý Úà Ù „ ''Ô — ¹ AÝ Úà Ù „ 0ԗ ç Ð ç Ð Ý à „Ù ˆ „ Ú Ý à Ù Ú „ ¹ &Ð — { å –ç —†– ã Ð Ý Úà Ù ''Ô ‡ †w 1Ý Úà Ù Ô — ã „ Ð ç Ý Ú eÝ à Ù ¹ ã w Ú æ { Ú Ô 1Ý „ à Ù Ð „ à Ý eÝ Úà Ù½× Ý Úà Ù iÝ Úà Ù 5Ý { ç Ú { ã Ú Õ š„AÝ Úà Ù Ý V Úà o „× Ð ˆ Ýç eÚ ‹ Ð æ w  —„ ã &Ð — { ç Ð Ý Úà Ù 'QÔ ‡ †w AÝ „ Úà Ù 0ԗ o ç Ð Ý Ý à „Ú–Ù ¹ ã w Ú Ôæ ~Ý à –Ù „ æ Ð { Ú Ý Ý à ÚUÙ × Ý à –Ù ãiÝ à GÙ ÚbÝ { ç Ú { Ú ß Ú Ú Õ |1Ý à –Ù Ý Q Úà × Ð ˆ „ Ý ˜8æÚ o „o ‹ Ð ç „ „ Ú Ù ¹ Ð Ý à „Ù × — – Ý Ù Ù‹ Ò – æ ˜zÐ Ý Úß ½ Ú× ¥zÐ Ý Úß Ù ˆ Ð eÝ Úà ½ Ú× ¥xÐ Ý Úà Ù Ý Ù Ù‹ ˆ „ CHAPTER 5. we would have .ú ý Aø ý š¦ ý ì vú ˆ6ý Aø “6ý › ù cù ÑG $# ñ › ` ë W ñ ëW ò ÷ ø  ú ý ñø ý ú ò ÷ ùø ¦ý ` úë Wý ñø ë Wý Xë W V ú ò ÷ ùøø Uˆx©˜› ý R kcS¯š6…ì v–ˆ6ˆAk“6˜› e“6ý cR ˜QS5ÑG $# When do you consume? If The value process is the smallest process with these properties. Î Î Î The discounted value of the portfolio is a supermartingale. Stopping Times and American Options „ Using w — Ð æ †– ã ’Ø Here. results in (Recall that ! If we had ): . 5.2 Value of Portfolio Hedging an American Option We get different answers for Now let us compute The value satisfies "  i  ú ý ñø  G“A“  ý or. the value of the European put. 79 ý ( . is the amount “consumed” at time .

By doing this.80 and the holder of the American option does not exercise.1. Define — Dä e u u Ý h× ã ÚÔ Ð Ý h Ú× u ‹ ‰  ã 6Y~bt The stopping time corresponds to “stopping the first time the value of the option agrees with its intrinsic value”. define ÓÒ Ð Ò Ð Ùo Ð ™ n Ð ™ d ي Ð ÕSÒ Ùo Ð S%¹ ¡Ï Ò æ Ð Ø × "¹ fö û C©  "  GUû  ù   ö i 3 …fhAi ö Q©þˆ6£©ç§Q ûû ý ì € T ú ø 4… ! € þ x TmD“5É"  GGû  ˆ  ù  Dmu†f4 © Q T ûû  Q ¢ ! e î 8 ý( Taý‰¥Q ì ú P ì  ¡ø ! ¤ ú ý xø“  ñ ú ý Aø ý › ñ úú ø ë ñø úú ø ë ñø ë bƒüéG’A¡  ö bƒüéU’Ak"› ýñ ð   “bú †¡ø ú ý ø ë Wý › ¢ð ú ý ñ  ø ë Wý ý Uˆñ ¢ k“6kx’Uˆ½¡3“6k› ö …ï . It is optimal for the owner of the American option agrees with its intrinsic value . so be the value functions defined for the American put with strike price 5.3 (A random time which is not a stopping time) In the same binomial model as in the previous example. be a probability space and let be a filtrawith the property that: if if ý› úú  1 ö bƒüéø ë Añø ë ›  ˆ ö ˆ ¯ ¬ 3œ ö ò ÷ ¶ œ ††û & ÷ 3 ú ø ¦ ë ` ú & ñ ø & fVù ë V8ùû ë& ² 3œ ö ƒüé°Y†ì G|A˜U› ù £ G #$ úú é ø & ñø & úú è ø & ñø & 1 úú ø ë ñø ë ö bƒüé%ÉAkU› ù ö bȃüé˜|A˜U›  2ö bƒÿ駒xk%› Ý D Ý Ú ×‰ ã 8 p Ý Ò ä ¡âÛÚ o EÐ ¡âÛ%u W~b†A@‡Ð ›âÛÚ C o £Ï i ú ý Aø ý › ö ý ñ Ù ØÏ Ï B å B B å )ß à è ß ß B)Û B )Û ó Ð ä Ý ¹ Ð ˆÜÛÚ Ð ä ç Ð ˆÜÛÚ Ý Ð ˆÜÛÚ Ý Ð ä – az Ý ç¹ Þ Ð ¡ÜÛÚ 7 A@9 Ð ¡ÜÛÚ 7 8 Ý p “ 7 Û 7 Û 7 Û ‰ ‰ ‰ ã ã ã 7 7 o ‹ é ö þ ë ÒًÒ؋ . he can ensure that for all . and . so there is a gap of size 1.1 (Stopping Time) Let tion. If the owner of the option does not exercise it at time one in the state . then the seller of the option can consume to close the gap. Therefore. œ In the previous example. then the seller can consume 1 at time 1. A stopping time is a random variable Example 5. It is an optimal exercise time.2 Consider the binomial model with . we have to exercise whenever its value Definition 5. Let . where is the value defined by the American algorithm in Section 5. Thereafter. he uses the usual hedging portfolio In the example. We note that à ¤ß We verify that is indeed a stopping time: Example 5.

Stopping Times and American Options where . but the set is not.e. random variable is given by 81 if if if 5. let The atoms of are Notation 5. 4 û x ì € T ú 4… þQ Ý ii Y ý f”Ai ö §—þø £¦G“xž ! mÁ ž 4 Definition 5. à ß if if å ß i. is a stochastic process adapted to this filtration. then is an -measurable random variable whose value at is given by 4 ú ¤ ‰ ì  ¡ø ! R ì — — ä à D–Và ß àã ä ß ß ä à ß Ò QfÐ å ß Ò …Gã Ò –¡„ã yB ä ß à ã Ð ä ß à Ù Ï £QfQä ç Ð ¡ÜÛÚ 7 ‰ Û ã î ðQ–ã Ý but o ÎÏ ä à ß B –¡„ã Ù B è Ø ÏÏ B û Gú`©þø X þ V R ô'ú`©þø”R ö ó ó R î 8 ý( Uý T è ó ä  ó Q ì þ S ¥ Ï ä – à ß ¡Gã The set is determined by time . Indeed.3 Information up to a Stopping Time Example 5.4 In the binomial model considered earlier.2 Let that be a stopping time. In other words. and is a stopping time with respect to the same filtration. R "ì Ý Ý Ú ‹‰ ã 8 p Ò ä e ku × ã ÚÔ Ð ku × %u Y~b†A@‡Ð 7 B)Û B )Û Ð ä Ý ä à ß ¹ Ð ¡âÛÚ 7 ‰ Û ã î –‰„ã Ð ä ç Ð ¡âÛÚ ä à ß Ý 7 ‰ Û ã –‰„ã Ð UxÐ ¡âÛÚ 7 ‰ Û ã îî –‰„ã ä – ä à ß Ý ß à Qã 7 Ý ç¹ Þ Ð ¡ÜÛÚ 7 Í 4 The collection of sets determined by is a -algebra.CHAPTER 5. We say that a set ä à à B –V–ã o yB ä ß à Ù £Ï ÃQ–ã ØÏ Ï y B à ß Ð ä Ý ¹ Ð ¡ÜÛÚ Ð ä ç Ð ¡ÜÛÚ Ý Ð ¡ÜÛÚ Ý Ð ä – Ux C We verify that is not a stopping time: ã à à VÐ Û ß à Ð Ò Q)ÛÛ Ò à ß B is determined by time provided H Ý ç¹ IÐ VÜÛÚ ‰ ‰ – PQ C C Û ‰ Û ã C Ûã C F × G Ø A@ Ð D F 8 t t b ‰ o C p  â o stops when the stock price reaches its minimum value.. is a filtration under .1 (Value of Stochastic Process at a Stopping Time) If is a probability space. This î 8 ý( U…UQ T ý ì . which we denote by .

. we obtain is a supermartingale. implies almost surely. then 82 almost surely.17 (Optional Sampling) Suppose that gale. then implies If is a martingale.e. i.2) I Ô æ H c ºÐ Ñ .4 considered earlier. 5. and in particular. then . If î 8 GT ý ì  ý Tý ò Q X ò G$ ò G$ ¹ C ú%"ì X øf ö W R R V 4 ú X %Rì  øf ba R V 4 X R Õ ò ` † Ž ò G $ ò  ò G$  ò G$  Example 5. we define probability measure. We compute é Ð ¡ÜÛÚ Ý × hi Ý Š Ú u In every case we have gotten (see Fig. Let and be bounded stopping times. å ß B fÛ ÐÑ ä à ß ä ß ß Ò –¡Gã Ò …Gã S ¥ Ï î 8 Tý U¡ì  ý T ý V 4 and for ò Q The atoms of are and . Under the risk-neutral is a martingale.5 In the example 5.. ) is a submartinsuch that . . . the discounted stock price process for all . there is a nonrandom number î 8 U˜xì  ý Tý T ý ò Q V 4 If Taking expectations.— I Ô Ð ›ÜÛÚ i h Ý¡ÜÛÚ qÛ £S × Ý eS i h pÛ £S æ H ® ® ® Ï® o × I o Ô c æ H gк Ñ Ð à Ý V Úà o × I o – å— ç æ —– Gå ˜ƒ { Ùo šå 0ԗ ¹  Ùo Ô { ß Ô æ H Ùo AÝ Q Úà o × I æ H Ùo o Ð ® Ð “âÛÚ e ¥Ï ®®® o × Ý S o ® ® ® Ï® o I à Ò Ý ¡ Úß o × I o I o Ô æ Ô æ H ß Ò Ý … Úß × o H Ð Ý ¡ Úß e S × Ôæ c º à Ð Ý ß… Úß e  × Ôæ c º S o H ÐÑ ® ® ® Ï® o I o H å ß ® ® ® Ï® o I — Ô c eS £ × æ dº o H u ÐÑ B )Û î ` D Ž ò ò G$  ö î ûúR `X ø G˜`ì †YŽòfG $  R W ò   "  V  "  4 " î 8 ý( axì  iQ T ý ýò (or Theorem 3. Therefore.

56 (4/5) S (T) = 1. Stopping Times and American Options 83 (16/25) S (HH) = 10.2: Illustrating the optional sampling theorem.40 1 (16/25) S2 (HT) = 2.60 1 (16/25) S2 (TT) = 0.CHAPTER 5.24 2 (4/5) S (H) = 6. .64 Figure 5.56 S =4 0 (16/25) S2 (TH) = 2.

84 .

he receives the payment . In particular û Û is an optimal exercise time. . The owner of an American derivative security can exercise at any time . and if he does. is an optimal exercise time. i (b) The discounted value process is the smallest supermartingale which satisfies almost surely.1 An American derivative security is a sequence of non-negative random variables such that each is -measurable.Chapter 6 Properties of American Derivative Securities 6. (d) The hedging portfolio is given by 85 ûù ð ûû %í"  U¥û  ù   ö i  úƒé  ›þ  U¥û  Q©þ3“6ˆˆÈè  ›þ  U¥û  Q©þ%ˆ6ˆñ ú ý û û ë ø ý ý ûû ë øë Wý ñ ð ú ý ûû ë øë Wý ú ƒé  ý þ  ¥Uû  ë ©þø “dý ˆÈè  ý þ  U¥û  ë ©þø “6ý ê ö ¥›þ  ¥Uû  Q©þ©†ï ûû ûû ëW ê ðú ëW ¦ b T U R úò ÷ 9r R ˜fVSùø ý ¡r ö i–cDA´¾½ ý ê… iQ R £ G #$ î ö ½ê ôö f4 4 (c) Any stopping time which satisfies i   t4 4 where the maximum is over all stopping times satisfying ¦  6 ý ` R Rúò ÷ ù …ì 1r R 6˜QSø £ G $# ý ˜Q°ø UƒÛö iê ú ò ÷ ù þ ýR û ý  ý  i 9 ¡uiê x ý r î 8 ý( aý ê ý R ˜fVS¨Q T úò ÷ ùø ý iê (a) The value of the security at time is ý ¡r ý …ì ý ¡r i î 8 ý( U¡sQ T ý r almost surely.1 The properties Definition 6.

i. If he does not. then the seller of the security can immediately consume þ i Let Define 6. a. be a sequence of non-negative random variables such that each is -measurable. then for every .2 Proofs of the Properties 86 and still maintain the hedge. ..e.s. "  ûû U¥û  ù   ö ûû GUû  ù   ö i  ýê  ý if`iò ý r  ¡u`ý ò î 8 ý( ˜iò ý 6˜VS¨Q T ý R ú ò ÷ ù ø ò Q  "  i  î 8 ý( aý T ö  ö ¶  ûý %iê ú ò ÷ ù ý R k@VSø Ä E GR w ¦ ý S‰ì kr ` R úò ÷ þ ý R R ˜VSùø £ G $# aƒû ¦ ý ì` y R r R ú ò ÷ y R 6˜fVSùø £ G $# ý `¦ë Wý R Rúò ÷ ù ² …ì 9Y“d…ì ` y r y R 6˜VSø £ G #$ ·G $# û ¶ë Wý R y ú ò ÷ ùø² VP“d…ì ` y r GR R ˜QSA­G #$ ö “6ý ê e“dý cR ˜QSø ë W Xë W V ú ò ÷ ù ë Wý “diê is a supermartingale. .s.20 If attain the maximum in the definition of to be the constant . Define also 4  i Lemma 2. we have is another process satisfying a. is a supermartingale.19 The process is also in ¦ ` W ú ÷ù ö ý…ì ëˆ6ýiê eë“W6ý VcR6˜ò@›°ø £ X ý ˆé G $# x '4 and Lemma 2. to be the set of all stopping times satisfying almost surely.18 ý r  ý ¡uiê Proof: Take ýé € ˆf¥4 Proof: Let x '4 Because Lemma 2. we have should exercise it. Then the owner of the derivative security (e) Suppose for some and . .. î 8 ý( T ý ê ý SkcS¨Q R ú ò ÷ ùø i i Ä G #$ E û ©¦ ý ì ` R r R 6˜QSø £ R ú ò ÷ ù ý ý …ì xr "  þ ý Uƒû w ú ò ÷ ù ôö GR ˜@›°ø Šý ê ý 4 ý ˆé î 8 ý( UxvQ T ý r ò ÷ ð ú ú ø ëW §—þ°¦ ý ì ` “6ý ê £ G $# fù Vù V§—þø ý ê ú øý §—þ©xr ö Q©þ©iê ú øý .

e. i.ò ÷ û½búƒé  ý þ  ¥ûUû  ë þ©ø ë“dý ê   µÈè  ý þ  U¥û  ë ©þø “6ý ê   ø @ù Vù ö ú û ÷ú ûû W ëW ú ûû ë ø¦ý `ë Wý ò ÷ Gý›þ  U¥û  Q©þ°S‰ì Dˆ6iê £ G $# @ù Vù ö ú ý þ  U¥û  ë —þø ý )Vú ý þ  ¥Uû  ë ©þø ý ê ûû ( ð ûû We prove the first equality. for each fixed we have and define recursively is a supermartingale.. Proof: The optional sampling theorem for the supermartingale CHAPTER 6. the proof of the second is similar. must be non-negative almost surely. The induction hypothesis is that .21 Define R û %ý ò úò ÷ ù kVSø   ö úò ÷ ù ý kVSø úò ÷ ù ý kVSø ¦ 6 ýò úò ÷ ù ý k@VSø ý Ä E GR w ¦ ý b¡ì 1Wò ` R úò ÷ þ ý R R ˜VSùø £ G #$ Ä aƒû w ý ` R ú ò ÷ ù þ ýG …ì ~r R R ˜VSø £ G #$ aE ƒRû  ¦  ý ö iê G $# ý `R úò ÷ ù …ì 1Wò R R ˜Q°ø £ ûýé € x "ˆf4 Y ý ò ý R k@VSø úò ÷ ù î 8 ý( ˜iò ý 6˜QS¨Q T ý R ú ò ÷ ùø We need to show that Proof: Then Set Since Therefore. Note first that ûú ý ûû ë øë Wý ú ý ûû ë øë W Gƒé  ›þ  UUû  Q©þU“diê ö ƒé  ›þ  U¥û  Q©þk“6ý ú û û þ ë W ö Èè  ý þ  ¥Uû  ë —þø ˆ6ý ó ú ûû ëW  Èè  ý þ  UUû  ë ©ø “dý ê ó û §ú ý þ  UUû  ë ©þø ý ê ö ú ý þ  UUû  ë ©þø ý ûû ûû úý ûû ë G›þ  UUû  c©þø T ù Uí"  UGû  ù  U“’i ð ûû Q € ó ýiê ö ý i ó û ý ii x iê ö ý ó ûú ñý ð ý ( ð øúò ÷ ùø ÷ ë Wý ñý G¥ýˆU†ïƒƒ'ý U˜QS­V“dˆ¥…ï ö “6ý ëW îê ö î ó ó û û ë ë W ˆÈ › UU c 3“6ˆ ú ý û û ë ø ý ñ ð ú ý ûû ë øë Wý ó û úúƒéé  ýVþþ  ¥Uûû  Q—þ©þø3“dý“ˆúÈèè  ý›þþ  UUûû  c©þ©þk“diêñ ö UVþ  U¥û  Q—þ©…ï ý û ø Wý › ûU¥ ëQ kë“6iê ð ûû ë øë Wý î 8 ý( Uiê ý R k@VS¨Q ý ƒ( T ý úò ÷ ùø û¦‚ý…ì ë“6ýiê eë“W6ý VcR ˜òf÷VSø £ G $#‡ðiê ý R ˜QSø qý ˜QSø ¦ ` W ú ù ý úò ÷ ù € úò ÷ ù X ý `ë Wý ò ÷ ð ý xì Dˆ6iê £ G #$ @ù Qù iê ö ¦  ý ö ƒ( Lemma 2. Properties of American Derivative Securities We proceed by induction on . Define implies for some 87 .

and using the portfolio .3 Compound European Derivative Securities In order to derive the optimal stopping time for an American derivative security. where each is -measurable. it will be useful to study compound European derivative securities. we will suppress these symbols. Thus a compound European derivative security is specified by the process . Superpose the hedges for the individual payments. Here is how we can hedge a short position in the ’th European derivative security. which are also interesting in their own right. start with wealth . In other words.88 will be fixed for the rest of the proof. different simple European derivative A compound European derivative security consists of securities (with the same underlying stock) expiring at times . For Since example. i. 4 . The value of European derivative security at time is given by À i  À  À "  and the hedging portfolio for that security is given by Thus. first make the payment and then use the portfolio T U ûù ð ûû %ŠÀ  UUû  û ú èø ë W GÈAk“6ý ê ö úƒéükë“6ý ê ÷ ÈAU“diê ÷   5ƒüék“di’ÈAk“diAø ö ø W ú èø ë Wý úú ø ë Wý ê ð ú èø ë Wý ê   ú ø ë Wý ÷ ú èø ë Wý ƒÿékˆ6iê iÈAk“diê   ÷  ý ú ý ñ ú ò ÷ ù ø ð ý ñ   ø ë W ˆý ñ ê ú ¢ ð ðÈ ø ë W Gˆb˜fVSCˆ½¡ø ú üék“di5y’¡¡Aøè  k“6ý ê ö ú ú ý ( ð ý êø ú ò ÷ ùø Uƒ)'iA¥k@QS…÷ ë Wý ú ñ ò ÷ ù ø ð ú è ø ë W ý ñ úƒ økë“W6ýiñð¡úÈèAøkë“Wdýˆ Uý“bú˜QSf’ÈAk“dˆxø ƒüéüék“d“¡ÈA%“6iêñ ö ú ø ë Wý ê ð ú èø ú ý ï ð ý ( ð ø ú ò ÷ ùø ÷ ú èø ë Wý ñ ý ký“ñU…—ƒ¡'ý U˜QS­iÈAk“dˆ˜…ï ö ÈAk“6ý ú èø ë W ó ó û úú ø ë Wý ÷ ú èø ë Wý ò ÷ µ5ƒüék“diê   iÈAk“diê   ø ù Qù ö ƒ)—ý ê ý ( ð ù ð ûû í"  G¥û  ù   ö ú ë X R 4 V 4 ï  U¥û  X 4 Vî Aø ûû ï V î ê X 4 úƒé  ›þ  U¥û  Q©þø “Xë 6V4W ý ’Èè  Vþ  U¥û  Q—þø ˆXë 6V4W ý ñ ý ûû ë ñ ð ú ý ûû ë i  úý ûû ë ý ú ý û û ë ëX ê ð ú ý û û ë ëX ƒé  ›þ  ¥Uû  c©þø “4 dVW ý ’Èè  Vþ  U¥û  Q—þø ˆ4 6VW ý ê ö ¥›þ  ¥Uû  Q©þø X 4 V ï ûû U¥û   ö î ûû GUû  ù  X U( 8 4 T 4 ( V  Q € D“i ý ï ÷ûûû …fUUC÷ ( v Q  i ¦  6 ù ÷ ²Ñ" ý …ì ` 4 ( 4 R ˜fVSø £ G $# ý ˜Q°ø ö X 4 V ý ê úò ÷ ù úò ÷ ù X & W Pcd ý ï Xë W ý ý ý V †f÷ Yˆ6ý V …ï ö …ï X 4 V îêî 8 4 ( „ î …ö ½ê 4 ì 4 ( ( 4 ( úý ûû ë G›þ  UUû  Q©þø ý ƒ( î À î ý …UQ ý ì À 8 U T ( 8 4 T 4 ( ( s Q ..e. the security that expires at time has payoff . At each time . we can ensure that at Hedging a short position (all payments). the last equation can be written simply as We compute 6. Hedging a short position (one payment). starting with wealth time we have wealth . the process is adapted to the filtration .

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 final 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 definition

|“ `˜

‡… „ ƒ š™˜ ˆ Gupy‚ sGGw š™˜ vUGw ‹ — –  •• ›~ s Gb…¥“ ˜ } —–   sw | “ ˜ Gbt• œw k w

n

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

which maximizes

“ pe} am ip l sj sw

.

o

n

f ge

‘ ú ò ÷ $ ( ˜fVSùø î ˆ ö î ê î ö î ê † ‡‰ †
4 4 R

‡… „ ˆ Guq0ƒ‚ { vw k w ‹ Œ~ s •  ˆ G…†py‚ am p ‡ „ƒ “ € ‰Š pk xyp m 0wuvs  ~s • p i ˆ£G†„py‚p“am p ‡… ƒ € ˆ~ s • sw ”˜ } |“ i

î

l k ˜” dj ‘  •  { Gf 2• f 6G“’`t'n

‰Š $pe p

( s
Q

Suppose you own a compound European derivative security
( X 4 (

. Compute

(

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

i ™e

{p x us • p Uzk yp m 0wvtrqe

î ½ê

n

8 4

G #$

d

V

#

#

#

#

8 T ý ï ë î R ý( a…kQ "

8 4

f k

“˜– • “ a™—”’
"

} |“ sw Ž˜

“m l k a6fi Gf vj

f ™– h

f k f ˜ £ “m l ‡… „ ƒ a£fi Gf ˜ f ˆ 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.

n

ˆ Gu0ƒ‚ ˆ Gu0ƒ‚ ‡… „ ‡… „ { “ t• ¨¢¡ w k sw ˜ Ÿ ®~ s £ ¢ ¬ w k ¬ w Ÿ ­~ s ¡
# # #

#

‡… „ ˆ Gu0ƒ‚ { ¢ ¬ w ¡¤¦ w k ¡ w Ÿ  ~ s • ‡… „ Gu0ƒ‚ ¢ ¬ w ¦ ¥¤¡ w ˜ ¡sw ˆ ˆ GuŸ py~ƒ ‚ s £ ‡… „ ‡… „ ˆ Gu0ƒ‚ • ¬ w k Gw ¬ w ˜ Gw ¬ ¬ © Gf 2• f 6G“’`t•  n l k ˜” dj ‘ 
a.s.
"

¡© w 2• ¡ w ˜ k
#

‡… „ G›q0ƒ‚ sGGw © vw k w ˆ ‹ –˜  †~ s š §™— bt• “ ˜ 
attains the maximum in the formula

n

• £ • £ “˜

# # #

§¢¡ w ¡¢ w ¢ “ ¦ ¥¤¡ w { A¢¡ w k sw ¡ ˆ G†py‚ ‡… „ ƒ

‡… „ {“ ˆ Gu0ƒ‚ Ga˜ k¡ w Ÿ ~s ‡… „ ˆ Gu0ƒ‚ ˜ sw Ÿ ~s ¡ ˆ‡… „ £Gu0ƒ‚ ˜¡ w Ÿ ~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:

Ÿ  ~ s • “ ˜
# Û

Therefore,

© qA¢¡ w k ¡ w

ˆ‡… „ ƒ £G†py‚

Ÿ ~ s • ¨¢¡ w ˜ ¡ w
# Û

ª©¡ w k2• ¡ w ˜ ˆ‡… „ £G›q0ƒ‚

Ÿ ~s • “ ˜
# Û

But we have defined

We have just shown that if

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

« « Y n  n

 žn

then

and

90

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

‡

‡ ‚ {  ·µ¤ ·’Œ~ s ¯ • ‹ ‚Á ¿¾ À ‡ ‚ ‹ — –½ ‹  ·µ¤ ·’Œ~ s G§b¼  £  ·µ¤ ’ §¯Œ~ s
This implies
is linear

{

‡ W¥º‚ Á §¾¼ À W¥º‚ ‡ G¿— b • ¯ –½
is linear

Jensen’s Inequality

Chapter 7

7.1 Jensen’s Inequality for Conditional Expectations

Lemma 1.23 If

‡ ‚ { » 9~ s £ » 9~ s ’ ’ » ‡ aº W¥º‚ ¹ ¸j • ¯ © l £© Gd• µ ‡ ‚ ‡ ‚ {  ·µ¤ ·’‹Œ~ s ¯ £  ¶µ¤ ’ §¯Œ~ s ‹ ‡ ‚ ´ ³ ›v¤ ’ ¤¯ ~ s ± ›± °h¯ s ² 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,

‡ ‚Á ‹ Ãu„  ¶µ¤ ¶’Æ~ s • Í ‹ s uµ¶„ ¤ ¶’Æ~ Å • ‡ ‚ ’  ·µ¤ ħ‹Œ~ s £  ¶µ¤ ’ §¯‹Œ~ s ¯

91

à „ u9º

 …•

‡ W%º‚ Á

, then

92

ϕ

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 Jensen’s 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 final payoff , i.e.,

Ë

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

•

‡ W¥º‚

Ë

‡ %ς U†q0ƒ‚ w ‡ ¥Ï‚ ‡… „ © w Ë ˆ ‹ —–  †~ s G”t•  i ‡w %ς i Ë

… Ì £‡ ‚ Ì Í• Ì Ë

‡

‡ ‚ { f ‚Ç ¯ • ‡ ‚ ÉÈ ‹ ÉÈ ‹  f ¤¦ Ê£f ·ÇŒ~ s ¯ £  f ¤¦ Ê6f Ç §¯Œ~ s ‡ aº¥‚ ‡ ‚ ‡ y‚tYaº¥‚ ÅÐ • ƒ „{ ‡ Ë Ì Ð ‡ {Ë ‡ – 0ƒt„º ‚ËÅÐ « Wº ‚ ‚ ‡ §Ì{ Ж Ð Ë • Ð Ë ƒ ‹ Ñ  © §Ì¡ÒÐ d ‡ ± ² 6© ¨ÌrCË s ´ ‹e
#

¸

Theorem 1.24 If

is a martingale and

is convex then

“m a6fi l ‡ f Ç ‚ Uj ¯

Ëi

‡ ¥Ï‚ f Ë

‡… „ ƒ ˆ Guy‚

‹ †~ s

“m l Ç a6fi §f sj
#

is a submartingale.

‡ ¥º‚ Î È E–

# Û

#

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
# # # †

ÖØf Ø f ¤¦ Ø £f

#

#

‡ ¥Ï‚ ˆ Gu0ƒ‚ ‡… „ ‡ ¥Ï‚ ˆ G›q0ƒ‚ ‡… „ { i Ëi ‹ 2~ s •  w Ë w ‹ — –w ›~ s Gb Û ‡ ¥Ï‚ ˆ Gu0ƒ‚ ‡… „ ‡ ¥Ï‚ ˆ G›q0ƒ‚ ‡… „ © i Ëi ‹ 2~ s «  w Ë w ‹ — –w ›~ s Gb ‡ ¥Ï‚ £G›q0ƒ‚ ˆ‡… „ ‹ ‡ { ¥Ï ‚ i £Gˇ ›q0ƒ‚ †~ s • ‡ ¥Ï‚ ¾Uu0ƒ‚ ˆ …i „ ˆ‡… „ ‹ Ù ›~ s g~ s «  w Ë w ‹ ›~ s ªÚvw ¤¦ i Ë i  ‡ ¥Ï‚ £G›q0ƒ‚ ˆ‡… „ ‡ ¥Ï‚ £G†py‚ ˆ‡… „ ƒ { vw ¤¦ i Ë i ‹ ›~ s « w Ë w  ‡… „ “m a6fi l ‡ f ¥Ï‚ Ë f ˆ Uuq0ƒ‚ j ‡ ¥Ï‚ ˆ G›q0ƒ‚ n ‡… „ • … „ © f Ëf Ï †pƒ ‡… „ G›q0ƒ‚ ¤¦ Ê6f ƒ „ × ~ s tË f ˆ ÉÈ £ Ó Ï …†pƒ ‡… „ ˆ G›q0ƒ‚ È £ Ö ÉÊ6f ƒ „ÓÔ ×Ë ~ s f ‡ %ς …uƒ £G›q0ƒ‚ ˆ‡… „ ‡ %ς ˆ‡… „ £Gu0ƒ‚ ¤¦ ÊÈ6f Ë ƒ ®~ s f É × • ¢ f ¤¦ Ê6f Ë } Ê6f | ÉÈ ÉÈ Ÿ ~s … „ … „ %ς uƒ Ï ›ƒ ÉÈ Ê6f Ë ƒ « $Õ£f ƒ ÔË ÖÉ È Ó
# # # P # # #

‡

So is a submartingale. Let optional sampling theorem implies

and

CHAPTER 7. Jensen’s Inequality

( λ x, g( λ x))

( λ x, λ g(x))

Figure 7.2: Proof of Cor. 2.25

be a stopping time satisfying

"

« n « Ì

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

x

(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. Define if if

Then

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

ƒ l Û © vv{ © © U†¡û {{ Ìj Ñ

n

Proof: Let The set

Ü { w 6f Ç ü ÉÈ ÿ „ þ f Ç x Õ£f G0wu s „ w Ç x f G0wu s ü þ ÉÈ ‹ ÉÈ ÿ  f ¤¦ Ê6f·ÇŒ~ s x Õ£f Gw0u s w Ç x f G0wu s ü È u „ þ u ÉÈ ¢ f ¤¦ ÊÈ£f Ç x ÉÊ6f ÿG0wvs w Ç x f G0wv Ÿs ~ s ü ¢ f ¤¦ w Ü } Õ£f | ŸÇ ~ s É ƒ t„ ýUl û « Uj gl û £ nUj f ¦ n ü f ¦ l û « Uj n úm l Ç a£fi Äf sj

be a martingale, and is in , so the set

é Õ ¾ ôö ëçç ã qq¡Ôá ãâ à é Õùçùç¾ææ ì ô 蔡Ôá êç ã ãâ à ôôIpòáà pá óõ ã â ñ è ê ã pçŽEÔá ÷ø ãâç £å ƒÕ`ªêà ì ðï í £G0î'ì è ê b”êEãÔá ãâ ê Õbbªêà í ì é íì ë ê ãæ bÞpá ­2YqAá'ß èç ãæ qƒÞpá å ä ã â à ƒ ‡ ‚ Ü { {{ Ì • ÞÛ © vs{ © © …bd © Ý } ¨Ý| w £f Ç
if if if if

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

n

Let be a stochastic process and let stopped process

be a stopping time. We denote by

“m Ü Ç a£fi l w £f sj

“m l Ç a6fi Gf vj

the

.

CHAPTER 7. Jensen’s Inequality

95

96 .

Chapter 8 Random Walks 8. ). Define The process Brownian motion.. and on each toss. It is shown in a Homework Problem that  ( ˆ & „ ( & Ö ¥ Ó f (& ü ¤ f 120 ¥ ü f ( ˆ & „ )'%#!û £ f ( & $ "   —  ¤ úm l  a£§f vf ¨j a£§f Gf ¢j ú m l ¤  8.1 First Passage Time Toss a coin infinitely many times. It is the first time the number of heads exceeds by one the number of tails. as is the probability of . the probability of is . The random variable is called the first passage time to 1. Then the sample space is the set of all infinite sequences of and . Assume the tosses are independent. 8.2 is almost surely finite and where 97 n ƒ É Êm { © s{v{ '¥¦© ü û © 'Ç p ü f ¤ p f ©Ì ü ú ¤ ƒ ‡ ‚ ü Ý 'Ç p ƒ£ © Ý © ¡ ü p üpÝ   if if ‡ ‚ {{{ ¡ ¡ ¡ svf©©¡ ü Ý ƒ´ ü n ü { l ™f ¤ ” Ì £ U“’` ü n ûj ‘ ¹ ¢ ¡ ¡ ú m§l W£f¨§f ¤ j   ‡ » » ‚   {{ vs{ © Ý © É É Ý ü Ý f ¤ . then . Define is a symmetric random walk (see Fig.g.1) Its analogue in continuous time is If never gets to 1 (e.

» Ì ² Ö A C #D@ È B@ Ù A Ü w £f wÚ ¢ ƒ (ˆ& „ (& { ³ Ö Ó ¥ Ì We consider fixed . so if if 3 ©£ ‡ ‚ ´ ü Ý n ¹ Ñ Ý ü are martingales.2: Illustrating two functions of  for every fixed (See Fig.2 for an illustration of the various functions involved). We want in (2.) Since and a stopped martingale is a martingale.1: The random walk process ¤ Ö ( ˆ& „ ¥ ( & Ó w £f Ü ¤ ( & 5 6 Ü  ~ s ü w £f 4~ s ü ƒ f Ï £ ü bf ¤ 9 2 ± ¡Ñ s ƒ  ü ™ú  ´ 'û ² . to let Furthermore. . .1). because we stop this martingale when it reaches 1. 8. so ´ ü n © ´ 2›³ n ( & « w £f Ü ¤ ( & « Ì ( ˆ „ & ¥ ( & Ó ƒ Ü « w 6f ¤ ´ 'û ² As . we have 7 8  Ü w £f f 2 eθ e−θ + 1 θ in part (2. (Take in part (i) of the Homework Problem and take (v). but we have to worry a bit that for some sequences .98 Mk k eθ e−θ + 2 1 θ Figure 8.1) Figure 8.

1): { ´ ü n © ´ 2›³ n {(& « wÚ if if A EC @ »ÈA@ Ù Ü w 6f Ö and Ö ( ( )& ( Ì ˆ& „ & ¢ ¥ ¥ ( & ü ( & Ü w £f Ó Ö Ó w £f Ü ( Q I G F RPHB( & ˆ& „ ¤ ( & ¥ 5 6 ( & « Ì ~s Ó w 6f Ü ¤ ( & § “A‘² f  .. We have just shown that these paths collectively have no probability.2). using the Bounded Convergence Theorem again.e.3 The moment generating function for 9 2 ¹ ´ n l ›³ Uj s úm l a£§f Gf j ¤ ƒ ü ´ nj { ¥l †³ U9W s ƒ ´ n © ü ¢ l ›³ Uj Ÿs ~ s ( ˆ & „ ( & { ( ˆ & ü Ø Ö ¥ Ó × ~s w Ì so we can let ƒ ‚  ©Ì Ñ U V ´ 'û ² Letting . We know there are paths of the symmetric random walk which never reach level 1. We want to find so that  8.3) Solution: Ì ü X Ì ü { ÒÖ ( ¥ „ ( ˆ & »‡( ˆ &‚ ¥ ˆ & X£ „ ( & X ˆ& „ £ ¥ ( ( & Ì Ó ü X X ‡ sƒ ‚ © Ì X Ñ Let be given. We therefore do not need the indicator (2. For all i.2). we obtain (2.2) . (In our infinite sample space . Random Walks 99 In addition. and using the Bounded Convergence Theorem.CHAPTER 8. to conclude ( ˆ & „ ( & © & « x § Tyu s Ö ¥ Ó ( & « Ì S w w ( ˆ & „ ( & ƒ S { ü"Ø x § Tw0u s Ö ¥ Ó ( & × ~ s w ˆ „ ƒ ü 7 Ü w £f Recall Equation (2. and we rewrite that equation as (2. in each path individually has zero probability). we have in (2.

so (3.» ƒ `» { » XX £ ƒ ` Xƒ ü £ £ 0 » ƒX ` ƒ $ X ü X £ ‡ £ˆ ‚ d X X É w n d ~ s ü w 4~ s X d ƒ © ³ X ³ Ì ©» X £ ƒX ` £ ƒ ü w 2~ s X  ƒ { ³ X ³ Ì©» X £ ƒX ` £ ƒ ü w 2~ s X X ( ( ˆ & „ ( & © Ö ¥ Ó ü © » X ƒX ` ƒ ü £ £ ˆ& 9 2 {Ì 9 2 ( ˆ & „ ( & © ( ˆ & ü Ø Ö ¥ Ó × ~s {» X £ ƒX ` £ ƒ ü (ˆ& ©» X £ ƒ » ƒX ` ƒ ³ »X £ ƒ £ ƒ © »X ³ ƒ X £ b £ ƒ © X £ bc³ X £ ƒ ‡ 0ƒ‚ ‡ 0ƒ‚ ³ X £ ³ » X £ ³ Ì ƒ ƒ (ˆ& ³ X ³ Ì ³ ƒ ü (ˆ& {» X £ ƒX ` Y a Recall Equation (2.1) . so we must have and Ì ‡ sƒ ‚ © Ì X Ñ so Recall that this becomes d We have computed the moment generating function for the first passage time to 1.3): w 8. Now . related by .4 Expectation of 9 2 We want Ì With We take the negative square root: 100 .

8. i.. {f ¤ » ¦ {{ È É È  f ¥¤ sv{ © £f ¤ © Ê£f ¤ ¤ { vw ¤ »  É  {{ sw ¦ ¤ ss{ © È w ¤ © ÊÈ w ¤ ¤ Én ƒ {{ ¥ vs{ © i© ü h © l ‡ vƒ ‚ { © Ì EX © » Ñ ©» » X X { ´ ü "~ s n ƒ © ¥l †³ U9W s ü ´ nj ƒ © l ü f ¤ U“’` •ü n ”ûj ‘ ƒ £ £ ƒ { ´ ü "~ s n ƒ `» ƒX ü ‡ É ˆ w n ‚ ~ s X ` X h c £ £ ü ” û ‘ ™f ¤ 'Ì £ Uj “’` •ü ¦n g ƒX ` £ ƒ ü w ˆ qpw 2~ s X n ‹†~ s ü ‹ ›~ s úm l W£§f §f ¤ j ‹ †~ s ü r s ‹ ›~ s Én £ » n É vn £ » n .6 General First Passage Times Define Then is the number of periods between the first arrival at level 1 and the first arrival at level 2.CHAPTER 8.e. The distribution of is the same as the distribution of (see Fig.. i. we can let to obtain Thus in summary: 8.e. this Markov property implies the Strong Markov property: random variable depending only on same random variable for any almost surely finite stopping time .5 The Strong Markov Property The random walk process is a Markov process. Random Walks e fX 101 in the equation Using the Monotone Convergence Theorem. 8.3). random variable depending only on same random variable In discrete time.

8. and thus „ … † ‡ (strong Markov property) (taking out what is known) τ1 τ 2 − τ1 not random τ2 k .‡ Ï È f £ ‚ . and payoff function Ï Ï ©f ‚ú ü f » » ¤ ü† ü … » © É ˆ É Éƒ ü © É ü '¥ ü ‚ 0 » g X £ d ‡ sƒ ‚ { © Ì X © Ñ » 0» X 0 » X £ ƒX ` £ £ ƒ $ ü Dw 2~ s X In general. The risk ‡ vƒ ‚ © Ì EX Ñ . £ ƒX ` ƒ $ ü ƒX ` £ ƒ €{ sw 4~ s ü Bpw 2~ s $ X X r ü ü ü ü ü  sw ¤¦ qpw tX†~ s ‹ r .3: General first passage times. Take expectations of both sides to get ‡ { 0 » X ƒX ` ƒ $ X £ ƒ £ s‚r w © ˆ ü sw r ‹ r  Pw Bpw tXŒ~ s ¤ sw X r ˆ ‹ r  Pw ¤ ¤ Pw Bpw tXŒ~ s sw X r r ˆ X‹ tŒ~ s X  Pw ¤¦ r r sw r Pw ˆ Bw p u xyPw ¤¦ r r w Bpw X sw wXv~ s r P For Consider the binomial model. .7 Example: Perpetual American Put 102 Mk Figure 8. with neutral probabilities are .

and stop the first time the price falls to 2. has the same distribution under as the stopping time in the previous section. and we stop when the stock price falls to 2.e. This observation leads to the following computations of value.. then steps will be required and the value of the option is # s s x { ¥“ » ‡ ‚ ’ƒ ƒ ƒ { “ ” ‡ ’ƒ ‚ £ •£ …„‚ ‡ ‡ Ï ¢ r C …„w ‚ ’ƒ Ÿ ˆ G~‡ uÈ ¥ 0ƒ‚ £ …s „ u ~s È r Cw £ r Cw ˆ g n Ws »ˆ ü ”ûj ‘ { l ¥ £ ™f ¤ U“’` •ü n ƒ ‡»‚ ‡ ‚ {{ ©¥ vs{ © “ '¦© ü o © ˆ “ É p É { c•ü p ¥ h ‡»‚ ‡ ‚ ‡ …„‚ “ { É ˆ p É { gü ¢ wfr DCe dC w ’ƒ ­~ s È ¥ £ Ÿ ™ ƒ £ o ‡»‚ ‡ ‚ ‡ …„‚ “ { “‰ ü » É { cü ¢ p C w ’ƒ Ÿ ~ s È ¥ £ {l ƒ » ‡ » ‚{ ü ‡ É {‰ ü ¢ p C w ‡ ’ƒ ‚ ­~ s È sƒ £ …„‚ ü $p C w | Ÿ } ˆ ƒ ü ” û j ‘ •ü ™f ¤ U“’` gÉ n £ # # ƒ ü $ wš | ¦© Ì gú n } ‘ ü ü gú # — » ü Ï ˜bú # ü }r C w| ü ü ü – ‘ Ï ‘ ü p ¥ ™ú g ¦ n Ï f where ‰  ¤ is a symmetric random walk under the risk-neutral measure. i.e. stop (exercise the put) as soon as the stock price falls to 2.CHAPTER 8. Here are some possible exercise rules: üú Ï down .e. so the value of this rule with this initial stock price is ‰  We define ƒ £ o In general. i.. This requires 2 down steps. Value of Rule 1: W# Value of Rule 2: This suggests that the optimal rule is Rule 1. Suppose Rule 0: Stop immediately. . Rule 1: Stop as soon as stock price falls to 2. at time Rule 2: Stop as soon as stock price falls to 1. denoted by . Random Walks W# 103 . i. if for some .. Suppose instead we start with . and the value of the put is if . at time Because the random walk is symmetric under .

then the initial price is at or below 2. so is . is the smallest process with properties (a) and (b). We must show: (a) (b) (c) is a supermartingale. and the value of the put is Proposed exercise rule: Exercise the put whenever the stock price is at or below 2. Since the put is perpetual. then and ƒ { « o ƒ © £ o for f Ï Ï Note: To simplify matters. with a possibly different . p¥ ü ú ‡ ‚ » ˆ ‡ { » ‚ p ¥» h ü { {“ ü » ˆ ‡ » ‚ » „ p ƒ‰ É { » É { “ ü Ø × » ˆ ‡ »p ‚ É » ’ „ ‡ » ‚ ’ » {„ “ { ‚ ü » p É {‡“{ ’ˆ ‚ p » É Y‡ ’ É p¥ É ÊÈ p ¥ ˆ ‡»‚ ‡ ‚ h ’ h ’ “ { gü p ¥ ‡ ˆ ‚ » Y‡ ‚ » ‡ ‚ É p É „ { É p¥ É ÊÈ p ¥ £ p ¥ Ï h ’ h ’ ü oh p ¥ gf ‡ Ï » ‚ » Շ Ï ‚ » „ { f É h É { ’ƒ ‡ f %ς ¥ h É { ’ƒ ü ‡ ¥Ï‚ ¢ f ¤¦ Ê6f Ÿ ~ s £ f ÉÈ ‡ ¥Ï‚ f ‡ …„‚ ¥ £ ‡ È p …„‚ È p¥ £ û Conjecture 1 The value of the perpetual put at time ƒ ƒ {{ © vv{ '¥ £ © £ © Ì © ü o © p ¥ is h h ’ƒ o # £ „ ‡ ‚ •ü p ¥ h h ƒ ú m§ a6f¨l ‡ f ¥Ï‚ j úm l a£f m‡ f ¥Ï‚ ‡ ‚ h k ‡ Ï § …„‚ f ‡ ’ƒ ¥Ï‚ © û jÈ f £ £ h f i h « o p¥ ‡ ‚ h p¥ ü p ¥ uú Ï . Just check that „ ‡ ‚ £ p ¥ ‡ » £ ‚ •ü p‡ ¥ ‚ h £ É ˆ p É { c•ü p ¥ “ h for This is straightforward. 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. In this case. We must show that ¥ £ o If . Proof: (b). .104 for some . the initial time is no this rule is given by different from any other time. we exercise If immediately. h for some . We must show that By assumption. The value of as we just defined it. we shall only consider initial stock prices of the form always of the form . Proof: (a).

Proof: (c). then ‡ ‚ ‡ ‚ ü p¥ h h ..1). „ ‡ ‚ { p ¥ £ ü ˆ p vt„ ³ ‚ p » ¥ £ ‰ ü ‡¥ ƒ ‡ ˆ …„‚ É » p ¥ Շ h ‰ ’ …„£ ‚ » ‰ ü „ É p ¥ £ ‡ ˆ Ê‚È p ¥ » £ ʇ ’ ‚ ü » É „ É p¥ É ÕÈ p ¥ ¥ ‡p ¥Ï‚ © ú ú Ç £ h û ‡ %ς £ { û f 9f Ç is like every other time. This concludes the proof of (b). then and ‡ ‚ “ gü ¥ „ „³ „ ɒ ¥ ü n n » #— „h » o» “ ü ‡ v0ƒ‚ ‰ { „ { “ { ü ’» Y‡ É ‚ ’» ‰ ‡ ˆ „ ‚ü » ’ ‚ » Շ h ’ É ph ¥ É ÊÈ p ¥ (7. i. h ’ £ „ There is a gap of size 1. With appropriate (but messy) conditioning . every time provided we let in (7.CHAPTER 8. We must show that Actually. the proof we give of (7.2) ’ i h h ’ ú m§l a£f¨Äf Ç ‡ ‚ j ‡ Ï …„‚ f ’ƒ © û i È f £ £ Ç f úm l Ç a£fi Gf sj Ï ú Ï ü p ¥ qú ƒ Ì « o If .1) h ’ h ’ p¥ ü ¥ ’ƒ ‡ ‚ ü p¥ h ƒ Ï « o f ¦ ü o If . then (a’) implies .e. so it will suffice to show (7. since the put is perpetual. .2) can be modified to prove (7. Let ‡ © È p¥ £ „ … ‚ ü p¥ £ „ ‡ ‚ ü p¥ h For . Suppose (a’) (b’) is some other process satisfying: is a supermartingale. on Suppose now that for some ƒ ü ” û j ’‘ ü {l £ o ™f Ϥ U“` l ¥ ™f U“’` ü n ü ”ûj ‘ Ï ¥ £ o ‰ £ •ú ‡ ¥Ï‚ ‡ p„‚ { ú ü È p ¥ £ £ WÇ ú h ƒ p¥ ü ú « o so if for some . Random Walks “ c 105 and There is a gap of size .2) be any number of the form .

not just at points of the form for integers . or later) that the stock price is 2 or less. In the example we worked out. we constructed so that is and if we exercise the put the first time the value of the put at time if the stock price at time is ( . 8. For except for . defined for all . In this case. then we can imagine the function . either (a) or (b) holds with equality. In such a situation. ‡ ¥Ï‚ f p¥ h h ‡ ¥Ï‚ ‡ Ï { ú ü¢È w h Ì 9 º ‡ Ï …„‚ ‡ ‚ {¢È ˆw ‡ » £ ‚ “ w ’ƒ ‡ Ÿ ‚ ~ s ü ‡ ¥Ï‚ É p É { gü p ¥ h ü ú ‡ ´ 6© Ì f Ï £ ‚ „ …‚ ‡ ‚ ‡ ‚ Ÿ Ÿ £ w ’ƒ ®~ s £ ¢ w Ç w ’ƒ ­~ s 9ú Ç û º ‡ a¥º‚ h h û i ©È ú m §f f a£¨l YÇ f ‡ ‚ j ’ƒ o ‡W¥º‚ ‡ a¥º‚ h h h û . This function should satisfy the conditions: d (c) At each . then (c) will be automatically satisfied.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. we have For This suggests the formula h We then have (see Fig. we need only verify properties (a) and (b). which gives the value of the perpetual American put when the stock price is .8 Difference Equation If we imagine stock prices which can fall at any point in .4): (a) (b) ‰ ³ º ³ `¥ for every º º „ ‡ W¥º‚ { “ « ³ ºÌ © £ © a£ ©w ü “ u v¢ „ ‡ ‚ ƒ { p¥ £ » ü p¥ he « o ‡ ‚ ‡ ‚ ƒ ¥ s “ ” p tü É ˆ p É { gü p ¥ e £ o h º ‡ » ‚ » aº ‚ » „‡ ‡ W¥º‚ ¢ w É ¥ É Ÿ ’ƒ £ h º Wº h p„‚ a¥º‚ h ‡ ‡ © i” È £ £ † £ (b) º aº ‚ ‡ © i © h ˆ º „Wº ‚ rÈ‚É ‡ ‚ † D‹ q º ‡ah º ‡ É ‚ Î £ £ (a) . 8.

Defining . This is an artifact of the discreteness of the binomial model.9 Distribution of First Passage Times we recall that We will use this moment generating function to obtain the distribution of . We first obtain the Taylor series expasion of as follows: ü Ì ™ú n W s Let be a symetric random walk under a probability measure . Random Walks 107 v(x) 5 (3. º x x x . with ¤ { s y º º » „ º » ü Ø ¥ ƒ É s¥ É × ƒ { ³ ƒ © l ™f ¤ ” Ì £ U“’` ü n ü ûj ‘ X ³ Ì ©» ’ƒ º s ³ ³ ‡ ‚ » Yaº ‚ » „‡ ü Ø ¥ º É ¥ É × ’ƒ X £ ƒX h ` £ ƒ ü w 2~ s X h X w 2~ s ³ º s ú m§l a£f¨Gf ¤ j ³ £ If . ‡ W%º‚ “ h .4: Graph of Check of condition (c): If or . in which an analogue of (a) or (b) holds with equality at every point. then (a) holds with equality. This artifact will disappear in the continuous model.2) 5 x Figure 8. then both (a) and (b) are strict.CHAPTER 8. ‰ ‰ 8. then (b) holds with equality: “ « º ³ Ì If .

{ 0 ¥ ƒ É Õm { l £ o ¥ ü G9W s É ˆ p » X §  p ü w 2~ s nj X » ‡sƒ ‚ o ƒ£ o ˆ » Ö ¥ Ó m p „ ¥ ü $ X £ o ¥ É p X ‡ »§  ‚ ƒX ü X z ƒ ü w 2~ s X » X £ ƒX ` £ º 0 {p ¥ » » ‡ vƒ ‚ o ƒ £ o ˆ » Ú É Ù m p „ ¥º ü $ § É p £ o ¥  ‡ É º 8sƒ £ o ‚  o ˆ » » Êm p p m‡ ¥ £ o ¥ ‚ É p Ú É Ù §  ü ú º $‡ ‚  o am p p Ì º } p | ƒ z ƒ ` § ƒ ü W%º‚ ‡ £ £ ‡ W¥º‚ z ü z »  ‡ msƒ ‚ ˆ »Ú Ù ü £ o ‚  8‡ É ¥ £ o ¥ É p  ‡ 8sƒ ‚ ˆ ‡ ‚ u£ } o É p} ¥ } { ‡ ‚ E} p ¥ } ~ƒ ü } ¥ {sv{ ‰ ¥ “ { {{ sv{ “ £ o ¥ £ o ¥ ‡ ‚ E} ¥ } ~ƒ ü ‡ Ì ‚ | } {p { ‡Wº 0ƒ‚ “ £ o ¥ sv{ “ } ‡ W¥º‚ } p z ‚ © p ˆ £ ‡ “ o ¥ E} p s¥{ v{ } “ ~ƒ ü } p | z { £ f C wr De p ™ {{ ‡ ‡ ‡ ‚ { Wº 0ƒ‚ sv{ W¥º‚ — ü Ì ·· © |ˆ £ — ü ·  z p “ ‡ W¥º‚ ‡ ‚ z » ˆ Wº 0ƒ‚ “‰ ‡ ‰ »ƒ ü ‡ Ì ‚· z © – » ˆ Wº £ 0ƒ‚ ƒ» ü W¥º‚ · z ‡ ‡ É ü ‡ Ì ‚ z © É º ƒ £` É ƒ ü a¥º‚  z ‡ Ì ü Ì z © £ £ ü z 108 But also. So we have The Taylor series expansion of is given by .

5: Reflection principle.CHAPTER 8.5. 8.6: Example with . (See Figures 8.) ¤ {{ © vv{ © “ '¥ ü o © 0 ¥ ¥ ‡ sƒ ‚ o $ ƒ £ o ˆ » Ö ƒ¥ Ó ü l ƒ £ o ¥ ü U9W s nj £ o ¥ É p » ƒ © É ü l ü U9W s nj ü o Figure 8.10 The Reflection Principle ƒ ˆ » ü ¥É p To count how many paths reach level 1 by time . Random Walks 109 Figure 8. 8.6. count all those for which double count all those for which . and ƒ £ o¥ “ € ˆ » £ gÉ p ¤ . Therefore.

ƒ ü ˆ » ˆ » { l £ „ qƒ É „ lž“ £ « É p j9W s ž€£ É ˆ » p 9W s „ l ƒ l“ j ¤ l“ £ ©agÉ p ¤ 9W s ¥ l j¤ » ‡ sƒ ‚ { 0 ¥ £o o ¥ $ ƒ£ o É ˆ p » Ú É Ù ü »  ‡ 8sƒ ‚ £ o ‚o ˆ »  8‡ ÚÉÙ ü o¥  ‡ ¥ ‡ ‚‡ ssƒ ‚ ‡ ‚ 8sƒ £ ‚  É p »   ¥ £ o ¥ £ o ¥ £ ¥ £ o ¥ o ¥ ‹ 8‡ “ £ o o ¥ ‚ o ˆ » Ú É Ù ü ‡ ‡ ‚ssƒ ‚ sƒ ‚ 8 sƒ £ ‚  É p » ‡ ‡  ¥ £ o ¥ £ o ¥ £ £ o o ‰ ‹ 8‡ “ £ £ o o ¥ ‚ o É ˆ p » Ú É Ù ü » 8‡ ‚ msƒ ‚ »  ‡ msƒ ‚  ‡ ¥ £ o ˆ » £ o msƒ £ o ‚ o ˆ » Ú Ù ‡ ü É ˆ » £ m‡ “ ‚ ˆ – £ o ¥ ƒÉ p £ ƒ o ¥ p» Ú É Ù – ü j j ƒ – ˆ l » £ qÉ p ƒ ¤ 9W s ƒ £ l £ ˆ ü » p ¤ 9ƒ W s ü ü p 9W s j j 9W s l £ ü £ ‹ £  l £ ürÉ ƒ p £ ‹ ƒ ¤ lž“ £ o ¥ « U9W s £ l £ o ¤ ¥ « G9W s ü l £ o ¥ ü U9W s nj nj nj ¥ £ o ˆ » ƒ p ˆ 9W s £ ü j» ü É ˆ¤ p » 9W s ü j ƒ ü rÉ p ¤¤ 9W s ü l £ o ¥ « U9W s j nj . . For 110 In other words.

we say that I and are equivalent. then P ɉ © ‚ ‹… ü Ç { Ç i  ƒ ‚ ·ÇŒ~ s ü 9~ s { ‹ Ç … i ‚ ‹  ‹Š…Œ~ s ü # exactly when Ws ‡ ‚ { ¦ d i © Ì ü  W s Ñ W# s # … †# ~s W# s Ì ü W# ‚ s ‡ ‚  … for every random variable W# for which ´ ³ ‚ ›Ä¤ ˆ… ¤ ~ s ‚ ‹ …  ˆ‡…Æ~ s ü †~ s W# s # W s s W# s W# s ‚ and is called the Radon-Nikodym derivative of with respect to I . I and are equivalent if and only if P P Ç … (Let and be related by the equation to see that (1.t.t. there is a nonegative random P variable such that (1.e.r.1) Remark 9. I . i.2) (1. 9.27 (Radon-Nikodym) Let I and P be two probability measures on a space .r. . and I is absolutely continuous with P P respect to .3) ‡ ¦© $z¹ ‚ Ì ü ‡ ‚  W# s © Ñ t¦ d i © W s Ì ü d „ ƒ ‚ Tg W# ‡ ‚ s  W s ü ‡ ‚  W# s ¦ Ñ  ‚ W# s .) 111 # If I and P are equivalent and is the Radon-Nikodym derivative of Radon-Nikodym derivative of I w. Under this assumption.2) and (1.2 If is absolutely continuous with respect to I . satisfying .1 Radon-Nikodym Theorem Theorem 1. is the (1.3) are the same. P w.1) implies the apparently stronger condition Remark 9.1 Equation (1. we also have ..Chapter 9 Pricing in terms of Market Probabilities: The Radon-Nikodym Theorem. P . Then we say that Assume that for every is absolutely continuous with respect to I .

so 9.2 Radon-Nikodym Martingales Let be the set of all sequences of coin tosses. and let . Assume Define the I -martingale P Proof: { W s f d ‚ ˆ… „ ƒ ü Ws # d or equivalently. ÷ £å — š ã â ç è —™ ã â ê è —™ ã â ç Õqùçà ÕvÕbùçà ÕvÕ` àê ” 9 ” # ’ ‡ ‚ ‡ ‚ © ¹ Ñ Ý i © Ì Ý W s © Ì Ý W s ‘ and for . Let correspond to probability for ” and for .1 (Radon-Nikodym Theorem) Let of length 2.112 Example 9. Let I be the market probability measure and let P be the risk-neutral probability measure.28 implies that if # is -measurable. then ‚ ‹ …  f ‹Š…Œ~ s ü †~ s {f‚ ü ‹  f ¤¦ t‚Œ~ s ü ÉÈ ‹ ‹ ÉÈ ‹  f ¤¦  Ê£f ¤¦ t‚Œ~ §s†~ s ü  f ¤¦ Ê6f t‚Œ~ s ƒ { Û © {vv{ © © Ì ü û ©  f ¤¦ ›‚Æ~ s f ‚ { ‹ •ü ‡ ݂ ‡ ‚ { ‡ Ý ‚ WW ss ü Ý ‚ . W# s # W# s so that I and P are equivalent. then for any ‚ ‹ { f ˆ‡…Æ~ s ‹ { ‹ “ f ¤¦ t‚Œ~ §s)Š…›~ s ‹ ‹ “ f ¤¦ ‚ˆ‡…Æ~ §s›~ s ‚ ‹  ˆ‡…Æ~ s # ü ü ü ü … †# ~s f ¦ … Lemma 2. Then . f ¦ d Ñ © … „ s f t‚Œ~ s ü  ‹ f ¦ … „ §s”~ s ‹ … „ ƒ … Note that Lemma 2.28 If is f ‚ We can check that is indeed a martingale: -measurable. The Radon-Nikodym derivative of with respect to I is P ë . the set of coin toss sequences correspond to probability for æí   “ ç ê ç è ê è ç è êí ê  qç bç q`0êbbvŽã Œ æ ‘ 9 Û ” ø è ˜Õ”b àê — ã â ê •– ñ ’ ñp ã â Þá á ð “ ÕpAáà ð ”  ç æí ¹ W# ê s .

it will be useful to introduce the following state price density process:  ž” Example 9.2 (Radon-Nikodym Theorem.CHAPTER 9. since Ñ p ¦ t Proof: Note first that ë å Êâ à ã { Û © vs{ © Ì ü û © f ‚ f {{ ‹ { p ¤¦ f‚‹Š…‹Œ~ s pƒp ‚ ü  p ¤¦ Š…Œ~ s û « o « Ì d Œ ’ “ ü Ws … „ ƒ ü W s ‹œ„ ƒ f ‚d … d „ s  p ¤¦ f ˆ‡…Æ~ s Tƒ ü W s  p ¤¦ f ‹Š…Œ~ s pƒp ‚ ‚ ‹ ‚ ‹ ã #  £” ¢ “ Œ ˆ‡… „ ¾U†q0ƒ‚ ƒ f‚ ‹ p ¦  p ¤¦ ‹Š…Œ~ s p É ‚ 㠔 “¡ # ã ü f¤ d # Ÿ  ” f ¦ å 㠄 Eƒ Ÿ  ” … Lemma 2. P W# values in the 2-period binomial model example.1 the values of the martingale We always have . continued) We show in Fig.28) W (Partial averaging) (Lemma 2.28) 9. then is -measurable. 9.1: Showing the are for I . So for any . .29 If is -measurable and f ‚ Figure 9.3 The State Price Density Process In order to express the value of a derivative security in terms of the market probabilities. not . The probabilities shown s . 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. we have .

4 Note that Now for an American derivative security Remark 9. : (Lemma 2.28) (Lemma 2. û f ge (a) 114 Remark 9. . 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 .29) .© o i pk6p ¤ £ p ‘ p ¤ ú am ip l p ‘ p ¤ j # # # ‹ ™ { p ¤¦ w k w Œ~ s e  GiG¥˜¦ w p ƒ ¤ ü ¤ ‡… „ ‡… „ ˆ G›q0ƒ‚ ™ G›q0ƒ‚ ü ‹ †~ s pƒp ‚ e  G¦G¥˜¦ w p  p ¤¦ w k w ‚ w ˆ Gu0ƒ‚ ™ G›q0ƒ‚ ‡… „ ‡… „ ü p‘ ‹ †~ s e  G¦G¥˜¦ w p  p ¤¦ w k w Û « o « Ì { vw k w Œ~ s sš GiG¥˜¦ w ü  ‹ ™ ˆ G›q0ƒ‚ ¤ sGGw ‡… „ ‹ †~ s š  ™ i˜¦ ¥ ü vw k w ‚ w ‡… „ ˆ G›q0ƒ‚ sGiG¥˜¦ w ü ú ‘ ‹ †~ s š  ™  vw k w úm l k a£fi Gf sj {p‘p ü ‹  p ¤¦ f e f Œ~ s¤ ü É É É ‹ ‹ ‹  p ¤¦  ÊÈ p ¤¦ f e f ¤ Œ~ ¨s¤ ›~ s ü  p ¤¦ ÊÈ p ‘ ÕÈ p ¤ Œ~ s ú am fp l p ‘ p ¤ j ‹  p ¤¦ f e f ¤ Œ~ s p ƒ ¤ ü ‡… „ Gu0ƒ‚ G›‚q0ƒ‚ p ¢ p ¤¦ f e f ‚ f ˆ ü Ÿ ~ s ‡ … p„ ‡… „ Gu0ƒ‚ p G›q0ƒ‚ ‡… „ ¢ p ¤¦ f e f ˆ ü p‘ Ÿ ~s p û « o « Ì ˆ Guf e 0ƒf ‚ ¤ Œ~ s ü ‡ … { „ ‹ ‚ Ÿ ­~ s ü f ‡… „ ˆ Gu0ƒ‚ Ÿ ~s ü ú ‘ is a martingale under I .3 More generally for (b) More generally for is a supermartingale under I . P .

00 2/3 ζ (ΗΤ) = 0. Fig.3 (Radon-NikodymTheorem. not P .72 2 115 1/3 S1 (T) = 2 ζ (Τ) = 0. is the value at time zero of a contract which pays $1 ‘ W# s Figure 9. The probabilities shown are for I .2. The state price values are shown in Fig.20 S (H) = 8 1 1/3 S =4 0 2/3 ζ0 = 1. ‡ Ý ‚ Ws ‡ Ý ‚ f ¤ í í  © f¤ ú am ip l p ‘ p ¤ j Ý û ¶ q ⠝ ì µ ­ à  s© . we have ª í We interpret by observing that at time if occurs. Example 9.6 1 2/3 ζ2(ΤΤ) = 0. . (1) If ß Ÿ ë ÷ ° ë å ¨ø Å9« Õ“çvæ ª bê væ ª ã ª ã ®ëø ãâ à ²â à è® ãâ ç à ²â ê Ä´Êq“çvæ í± ª ”“çv±à í æ ª Õùç¾±à í æ ª ãâ è® ëø å °Tø Ńã vå « ÕŽê 6æ ª ³”bê væ ª Êbê 6æ ª ã â ç ±í à ² â ê±í à ã â à ±í ±í ±í é¯ å « å¯ Õbªêà æ ª ® ë ø ã Tø ž¡® ë ãâ © ª bbªà æ © é ¯ å ã vå ©® é ë å ã vå « ® é ë å â ê ê å « å ® ë ø 㯠â ê â ê s ¯ëø å ÕbbŸ ªêà í bbªêà í ë ÷ ° ë å ø ™ ë ¦å « å « å ã ª ã ­ ë® ãâ ç °‹Êq“çà ª Õbùçà ª Õ© `ªêà ª ãâ ê ãâç å ãâ ê ë ­ å « ø ã â ê âí ê Åø ë iå ã så ¬s ëø å í Քb àê ª bbªêà è vå í b” àê .CHAPTER 9. the value is ™ For a European Call with strike price 5.2: Showing the state price values ã æ ã ¨ ‘ § f¤ í . Pricing in terms of Market Probabilities ζ (ΗΗ) = 1. 9. We compute the value of the put under various stopping times : ¯ ‘ ‘ ë ÷ ¯ ë å ã® ™ ë å « ² é ° Ä4« « å ë® í å Õqíç ùç'Ebùç'Õè é Õ`ªê'¡Õ”bê 'ß ãâ àß ãâ ê àß ãâç àß ãâ ê à (0) Stop immediately: value is 1.36 S2 (TT) = 1 (c) is the smallest process having properties (a) and (b).3 shows the values of .72 2 S2 (HT) = 4 S2 (TH) = 4 ζ (ΤΗ) = 0. 9. Compare with the risk-neutral pricing formulas: Now consider an American put with strike price 5 and expiration time 2. expiration time 2.44 2 S2 (HH) = 16 1/3 ζ1(Η) = 1. Recall that . continued) We illustrate the use of the valuation formulas for European and American derivative securities in terms of market probabilities.

Also let É tosses. The probabilities shown are for « å ¯ s .S2(TT)) = 1. 9. . and assume P equivalent.S2(HH))+= 0 + (5 .S2(HT)) = 0. ÈÆ Æ Ù È É Ù )Ì Ê Ö Õ Ê Ô Ó Ò Ø×c£©8†Ñ Ð# Ê Ö Ú Ê Ô Ó Ò ØH£©82Ñ #Ð Ñ Ð# Let be the risk-neutral probability measure: Ï Î ¾  Á À Í Ë ¾ 8Va'D©³Ä ½ ¾ ¸Ë ¾ Ä È Ê H¾ Ì Æ ¾ ½ Ë ¾  Á ¿´©V•£À ½ ¾ ‹Ë ¾ Ä Ê ¬¾ È Let are be the set of sequences of -measurable.S2(HT)) = 1 + ζ2 (HT) (5 .S (TH))+= 1 2 + ζ (TH) (5 .S1(T)) = 3 ζ (T) (5 .S (H)) = 0 1 += 0 ζ (H) (5 .72 2 2 2/3 + (5 . the value is We see that (1) is optimal stopping rule. and let ÷ — °¸rs ëø å ë® ã ø « í í ‘ « ‘ ² é Š°°·« å¯ ë® « ‘ ² é Š°°·« ë® £ „ … ‚ Figure 9.S (TH)) = 0.S (T))+ = 1. where for each . not . W# í ‡ Ï È f f¤ í ‘ for an American put.116 (5 . Define Î ÓÍ ìë·Ñ Ð Î ÓÍ áêÃÑ #Ð Ê Î Ó £áêÍ é ¹vçèÓ戻å䕩8†Ñ Ð Ö Ó Ò Ï Ê €ã ¾ Ç 2Ê fi¾ Ó 2Ñ Ð# â Ú Þ ¾ Ô à Æ ¹ ç Ó ~îíæ Æ #¾ ÈÉ Ì Ê ã ¾ â Õ Ê Ô à ¾ Þ €¦·Ç •cáf¦ßÓ ·Ñ Ð# º Ü Å Ü aÃaÝÛ and for .S (H)) 1 1 1/3 (5-S 0)+=1 + ζ0 (5-S 0) =1 2/3 + (5 .3: Showing the values P I . Then I and P are .44 (2) If we stop at time 2.4 Stochastic Volatility Binomial Model ¾ Â Æ ¾ Ä Æ ¾ )¦#ri#¿½ Å È ¾ Ä ¼ ¾  Á À ¼ ¾ ½ ¼ ³€H)Ãcœ¿a2» º ¾ Ç ¹ Ñ Ð# Let I be the market probability measure.S (HH))+= 0 2 ζ2 (HH) (5 .S2(TT)) = 4 + ζ2 (TT) (5 .72 (5 .80 1 1 1/3 2/3 1/3 + (5 .

and consists of Ï À Ë º •ÃiÆ Ï Ï V)Ï Æ » Ê Å Æ Î ¾ ý ¾ ù Ë ¾ üÍ Î ¾  Á À Í Á Ô à ¾ ý ¾ ù Ê Ô à ¾ c2£8V )·¸ðÿ)V)€|þc|fif)·c ¦ðü û ú Ö ¾ ù ¦¾)TÝ¨Ò Ù ü and the following processes are martingales under I : P Ï û Ö ¾ ü i¾ú Eð8¾ Ò and We thus have the following pricing formulas: Å ¾ Ø£ Simple European derivative security with payoff at time : American derivative security : The usual hedging portfolio formulas still work. The self-financing value process (wealth process) . Æ Ù Ï ' & $ (%%# ã Ç â #  # €ï Ð Þ ¥ ø ¥  !  ") À ø  ¥ û ¦¾  ¥ Ù ã Ç dؾ aï Ð ¥ ø â ¾ £ Þ ¥  À ø   ¾ Ç  # ¡ ¾ ¢Ãü #  Ç ñ ¾ Ø£ ¾ ¨ ñ ñ and Ù ú À ¨ ©ï #Ð ï Ð# ø   ' & $ (%%# !  " ¥ ñ Ñ Ð# Then the following processes are martingales under ú : Ï Ï Ï V)Ï ¥ §ñ ÔÙ €ô Ê Ê ¾ Ù û Ö ¾ ý ¦¾ú o )¾ Ê Ê û i¾ ¥ ¦¤ Ù ¡ ¾ ¢ ý ¥ ¤ ø ¾ Ò À ñ   Ô ô ¾ õ8÷ö ¾ Note that is -measurable. Ï º ¦Æ Ï Ï Ï )VÏ Ï Ï 8VÏ Æ À Ê Å Æ Ô ô #ò2£Tõ8¾ ñ Î Ô ô ¾  Á À Í Ê s€m)€'Dó¬¾ ñ ñ . the non-random initial wealth. Pricing in terms of Market Probabilities º iÆ Æ À Æ » Ê Å Æ ã ¾ #Vc·‹ym2Ç â é Þ Ð Ê ðï cؾ é û Ö ¾  i¾ú E4Ò 117 We define the money market price process as follows: Æ À #òÊ Ù ñ We then define the state price process to be º iÆ Æ » Ê Å c2£Æ ¾ é À ñ Ê ¾ ø As before the portfolio process is .CHAPTER 9.

and since is -measurable.118 9. and the rest follows from the “standard machine”.-algebra of . The Radon-Nikodym theorem implies that there exists a -measurable random variable such that Ñ Ð# »cÊ'ά58ÍðÑ Ð# ÑÐ Æ 1 ç 5 @v9æ Ñ Ð# Ñ ¿½ Ð é » Ê Î 5Í c'¬8%0 é i. define two probability measures ü if for some .5 Another Applicaton of the Radon-Nikodym Theorem Let be a probability space. This shows that has the “partial averaging” property. In other words. The existence of conditional expectations is a consequence of the Radon-Nikodym theorem. If and . 1 1 é Ï 1 ç 5 @v9æ A 0 G4½ ü F Ñ ¿½ Ð ¹ 0 C é Ê Ð ½ cÑ ¿òF E C %DÊ E C BØ8ÃÑ Ð# RÊ Î 5 Í 0 ½ ü E 4è†%C ¹ C » Ê Î 5 Í cØ6·Ñ Ð 1 ç Pò5 1 ç Qv5 1 é 1 F Whenever is a -measurable random variable.e. it is the conditional expectation (under the probability measure ) of given . then . and let be a non-negative random variable with . We construct the conditional expectation (under ) of given .. the measure is absolutely continuous with respect to the measure . Let be a sub. so . we have 0 ü Ç Ï A 1 ç 5 B@v9æ 1 ç 5 @v9æ 2 0 ½ ü E C Ê Î 5Í 4ˆ2%D'Ø8ÃÑ #Ð Î 5 ¬8Í 0 Ê Î 5Í 7Ø6·Ñ Ð 1 À Ê 0 4½ ü ¹ 3 Î 0 Æ Ç Æ ¹ 4ðs~…Í 1 E H Ê IóF 1 . On . this is just the definition of .

1 Regardless of the portfolio used by the agent.Chapter 10 Capital Asset Pricing 10. i.e. Æ Ù ü Ê œc ú ø ïÐ then there is a portfolio which starts with initial wealth and produces at time . so P c c c .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. Remark 10.1 An Optimization Problem ü Consider an agent who has initial wealth as to maximize Here. ú ü Remarks 10. just regard as a simple European derivative security paying off at time . there is a hedging portfolio which produces . Ù ü Ê ú ü ú ø ïÐ Ù ø Ù ü Ï ú ü W US XVTï Ð Ù and wants to invest in the stock and money markets so is a martingale under I . To see this. and starting from this value..2 If is any random variable satisfying (BC). we wish to ¹ 119 Î ÓÍ Ð Î Î ÓÍ c W US áê·Ñ sáërpXViÍ $ Ó h Maximize Ï ü Ê œc ú ïÐ Subject to Ù c W US g4feï Ð Maximize ø º c dÊ Ù ú ü ü c 7Ê º ú Î £ a 4b8Í ü û Y¾ Ö ¾ ü ¦`Eð8¾ Ò Remark 10. Then is its value at time 0.1 and 10. (BC) stands for “Budget Constraint”.

Ï Ù ü Ê À ƒ ‡ ˆÊ Ù ü Ê Î ¾ ÓÍ ')ìë·Ñ Ð Ô û f¦¾ sr ú h ƒ À Î ¾ Ó Ï )áêÍ À ø ƒ Ê ¾ x t Equation (1.1’) implies ú Ô û  ¦¾ Ï Î† À wÛ DÍ Ï Ï V)Ï Ï Ù ü ʋÎT¾áêÍ·Ñ Ð ¾ VáêÍ Ó x tÎ ¾ Ó ø ú ø Ï Î )À DÍ † À Æ ú Û TiÆ Æ À Ê Å Æ Î ¾ ÓÍ Ð Î ¾ Ó oò·£8Váê·Ñ sVáêÍ Ï » Ê g£Î Ï Ï ))Ï x€ t Æ Æ Ô x tÍ ‚ and Æ … )Æ Ï Ï ))Ï Ï Ï V)Ï Æ#ÀòÊ2Å´ÆmÎ € Æ x t Æ Ô f‚ x tÍ ¾ Ï Ï )VÏ ¾ t „ ƒIÊ'Î t „ Æ Ô Í x€ t Æ x t ' „ „ € tÆ Ï Ï )VÏ Î tÍ ‚ Ï Ï V)Ï € t Æ Î Í t | x€ yt Æ x yt Í Ï ú Ô û fi¾ »cÊwGüvË Î)¾ìëÍ·Ñ )¾ )áêÍ Ó Ð tÎ ¾ Ó ø Ô û fi¾ sr Î)¾áëÍ·Ñ Ðs)¾ Ó Î W US t XViÍ sr Ï Î s r ÓÍ c Ê s #vê€d†vr t Æ Ï Ï )VÏ Æ Î r ÓÍ c Ê m8pê€dr Ï Ï ))Ï Æ r Ó Æ Ô Tpþ)©Ó ú ø Æ Î Ô ÓÍ c Ê t 8i©ê€uáÔ t Ï Ù 120 Subject to » cÊ ü Ë Î ÓÍ Ð Î ÓÍ c Î Ó ‹áê·Ñ Pßê€qáëÍ ¹ $ Ó h There are sequences in .2) become ¹ Ó then there is a number such that solve the problem . Adopt the notation (1.2) (1. Call them sr Ó pþÆ We can thus restate the problem as: Maximize h Subject to h In order to solve this problem we use: Theorem 1.2’) we get sr h ƒ Ê Î ¾ ÓÍ I'Táê·Ñ Ð x t À ¾ For our problem. (1.30 (Lagrange Multiplier) If Maxmize Æ )Ô Æ Ô Subject to Æ » Ê g'Î Ï Ï ))Ï Æ RÔ Plugging this into (1.1) ƒ ú Û .1) and (1.

3). then : .4). we have and let Let be any random variable satisfying is maximized at and define is given by (1. Capital Asset Pricing solves the problem Maximize Subject to c W US pXVTï Ð Theorem 1. solves the problem (1.Ï c W US Ð c W US x eXVTï €Ü eXVeï Ð ú ü Ë c W US œXVeï Ð Æ À #aË Ù c W US Ð Ü Î Í x eXVTï €oc ø %ï Ð À ú Ï À €Ë ˜ Ù ø ü ˜ – W US QXVdÜ ™Ù ú ø Ï Ù Ù ü ú ø ü Ê ú ø ü – c Ë c W U —PœpXVS From (1.3) 121 c .4) (1.. Thus we have shown that if CHAPTER 10.5) (1.31 If é Proof: Fix » aä We maximize then over » aä  and so Taking expectations. (1.5) we have x c Í %ï Ð Ê Î 'oc Ï » €ä é æ Æ » f€ä æ Æ À W US Ê Í  t ÷o•Ë À XV7'Î x t '•Ü é Ԕ Ê é W U t Ë t XVS x t The function  Ï † “ Ð ç t fa¼ r t ò'Î t Í ’†  æ Æ » Ë Ê À Ê Æ À é t Ï é ‡ ‰ » ‘gÊ Ë é À t Ê 'Î Í † t qy t W US Ê Í t Ë t XV7'Î t ' x c x c Ê ü Ï ú ø Ù x c Æ Ù ü Ê Î 'dc ú ø Í %ï Ð Ï ú iÆ Û Ï Ï V)Ï Æ À Ê Å #ò·'Æ Î ¾ Ó VáêÍ Ù ü ú ø Ê ¾ x c x t Therefore.e. i.

i. 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 ü and the optimal portfolio is given by ø The optimal 122 is h f iøgÊ ú ú ü ú ü Since Ù û ¦¾ú Ö ¾ ü ¾ Ò so is a martingale under I .Ï Î Ú Æ ¾ Ó #¬s#'DÆ Ï Ï V)Ï Ï Ï ))Ï Æ Ô ÓÍ Ô à ¾ ý Ë R£ëEfif‹ Î Õ Æ ¾ Ó io|DÆ Ï Ï ))Ï Ï Ï V)Ï Æ Ô ÓÍ Ô à ¾ RêEf¦ ý m & o|DÆ k ¾ Ó m k ¾ Ó h Æ Ô Ój Ô à if )£›( ¦¾ ø Ë nl %'DÆ ¾ Æ Ù Ï Ï V)Ï ø ü h Æ Ô Ój Ô à if R£›f¦¾ ø Ê 'Î ¾ DÆ Ó Ï Ï )VÏ Æ Ô ëÍ ¾ ù Ó Ê ¾ ¬ðü ú ü ú ø Æ º ÷iÆ Æ c·£Æ » Ê Å Ù Ï ü Ê ã ¾ €m4Ç â Þ Ð Ê ¾ ü Ãï gØðm¾ ø Ù Ï ú ü Ê ú ü ú ø ü W US XVTï Ð Ù In summary..e. . we have P .

e. “ •n uü Ð r ¹ p 2 Ç t t is defined for every ... 11. (the -algebra of (See Fig. defined – .e. We now consider a general random variable defined on the probability space . Recall that: is a -algebra of subsets of . is often called the Law of Î Î Ð Í v Æ Ð sH“ pØv2“ …Í Æ Î Ð 8H“ …Í v Qç a 9æ Ô €ô {Î a |œ8Í f y Ô õô ü z Ð Ê Î a sIÑ cœ8Í f } f y ü Thus any random variable by induces a measure on the measurable space Î Ð H“ pÍ v ü “ •n'p Ð r ¹ i.1) ü is a random variable if and only if is a function from to 2 Î Ð H“ …Í v uç Æ Ç ç Ö a ç Î Ó î%22@|ëÍ a A function is a random variable if and only if for every Borel subsets of I ). . Ç ç %c5 Î 5Í ¬8·Ñ Ð ¹ Ç Î s Æ Ç Æ ¹ £vÃsH…Í Ç P I is a probability measure on 11.Chapter 11 General Random Variables 11.2 Density of a Random Variable Î  Æ »Þ r Ð •)t~“ ¿p 123 Ï Î ÐÍ v ç a H“ …œQŽ9æ ½ t 'Î t Í f f €  %C  Ê Î a £œ6Í f y ü The density of (if it exists) is a function such that ü f y Ç ç Î a %©†8Í ü where the probabiliy on the right is defined since in Williams’ book this is denoted by . the set R ü A Ó Ò Px8vR Ê †8Í Î a Ô õô ü Ö a R Ê 2wç ü )Ò Ô €ô “ €õqü Ð r ¹ p .1 Law of a Random Variable Thus far we have considered only random variables whose domain and range are discrete. i.

‚ Æ ½ t 'Î t Í Ö a @ç Î a œ8Í f  € C ü Ò )†Ñ f “ a†a Ð … y RÊ Ê Ê Î ü ‹ÿÍ € H ‡ï Ð Î Proof: (Sketch). Now use the “standard machine” to get the equations for general .32 (Expectation of a function of ) Let be given. Thus has a density if and only if is absolutely continuous with respect to Lebesgue measure.1: Illustrating a real-valued random variable We then write where the integral is with respect to the Lebesgue measure on I . which means that whenever has Lebesgue measure zero.3 Expectation “ €~“ #w‚ Ð r Ð p ü Theorem 3. ü {X ε B} Ω Figure 11. then these equations are f y v ç a Î ÓÍ Ð ½ Î Î ÓÍ ü áê·Ñ ¿'s£êaÿÍ f  Ï ½ t 'Î t Í Î tÍ ü Æ f ½ t sÎ t Í Ï ƒ y ½ X'Î t Í ‚ „ XC » gÊ 2wç Ö a f  vÎ t Í ‚ ƒ„ C f  Ê D'Î t Í ‚ ¹ C ü Ò )†Ñ Ð RÊ Ê Ê f y X½ Î üÍ ‚ ‹ÿ÷‘ï Ð H Ê t Í € I'Î t Í ‚ f y . then Î Ð H“ pÍ 11. R is the Radon-Nikodym derivative of with respect to the Lebesgue measure.124 X R B . If for some . Then which are true by definition.

11.Y) C y { (X.CHAPTER 11.2: Two real-valued random variables . General Random Variables 125 (X.2) called the joint law of . defined by measure on Î s Æ Ç Æ ¹ £vÃsHpÍ induce a that satisfies Œ ƒ„ f ‹ ¢½ t ÷¿‹ Æ t Í kˆ v¿‹ Æ t  Å yC ½ Î Î Í C Î ÓÍ Ð ½ Î Î ÓÍ F Æ Î ÓÍ üÍ áê·Ñ ¿'sáêð8£êaÿ Å Î ‹ d)Æ Œ kˆ f 4½ ¿‹ Æ  Å ˆ C ƒ„ y Î C y tÍ tÍ F Æ ü We compute the expectation of a function of ¹ C kˆ f y is the Radon-Nikodym derivative of with respect to the Lebesgue measure (area) on in a manner analogous to the univariate case: r “Ð r “ …HQç ÐÍ v £ ‰æ Î  Æ »Þ •)tcr f ‹ ¢½ t '¿‹ Æ t Í kˆ  ½ Î r “ #p Ð kˆ f  RÊ Ê Ê Š C yC Î F œÆ Ê Î £ '4iÍ üÍ Å ÿ 2ï Ð Î F œÆ kˆ f y ü ÿÍ The joint density of is a function Î r “ pÍ Qó‰æ Ð v ç £ Î F œvÆ ü ÿÍ Ö £ ç Î F ·QœvÆ ü Ò Ð ÿÍ †Ñ cR Ê 2iÍ Î £ “ •n¹ Ð r kˆ f y Î r “ pÍ v Ð F Æ ü . Then Let (see Fig. 11. kˆ f  .4 Two random variables F vÆ ü be two random variables defined on the space .Y)ε C} Ω x F Æ ü Ï Î Ï Figure 11.

be given.4) (6.Ï Î Ð Í v ç a H“ p~‘ó9æ f ‹ ¢½ t ÷¿‹ Æ t Í kˆ Î t ÷‚ „ 4G%C ½ Î  Í ƒ C € Ê ‹ ½ Î ‹ ¢'¿8Í ˆ vd6f‚ “C Î ‹Í € Ï Î Î Ó s£êÍ F Í ‚ Ê Î ÓÍ ã â Î ü Í ‚Þ fd'£êþF )‹ÿ÷Ãï Ð “ a~“ ¬g‚ Ð r Ð p F kˆ f  Ó ã Î F Í 2â Î ü Í ‚Þ B2i ’)‹ÿ÷Ãï Ð Î F 4Æ ü Í Ï kˆ f  „ XC R Ê ¿8Í ˆ  ƒ Î Î ‹ t ½ ¿‹ Æ t Í where Æ ‹ ½ Î ‹ ³™'d6Í ˆ  € C Ê Ê Ê Ê Ê f ‹¢½ t ½÷¿‹ Æ t Í kˆ  ƒ„ C € C Î f Î †aŽI“ …Í kˆ  Ð y Ö2aPI“ ÐîÎ2vÆ ÿÍ †Ñ Ð ç F ü Ò a ç F Ò @˜†Ñ Ð Ö Î a œ8Í ˆ y Suppose Î F œvÆ ü ÿÍ 11. i.. there is a function ( depending on ) such that ‚ ã â Î ü ‚Þ R Ê ©F )‹ÿÍ Ãï Ð ‚ ¿8Í ‚ Î ‹ How do we determine ? We can characterize using partial averaging: Recall that .5 Marginal Density 126 has joint density . Let be given. Then for some (6.e.6 Conditional Expectation has joint density . Then the following are equivalent characterizations of : . is the (marginal) density for F Î ‹ ¿8Í ˆ  Suppose 11.3) (6.1) Ö Æ Î Ð 8H“ …Í v Qç a 9æ Î ¿‹ Æ Œ ˆ ƒ„ kˆ f 4'Î ÷v¿8Í H ˆC C y ½ tÍ ‚Î ‹ tÍ € Ê ¹ Î ‹ ½ d™…Í ˆ ’d8Í q¿8Í yÎ ‹ ‚Î ‹ € ¹ ƒ H „ XC Æ Î Ð mH“ …Í v ‘ç a 9æ Ñ ¿½ ‹ÿ÷v4Í Ð Î üÍ ‚ Î F € H C Ê Ñ ¿'œiÍ Ð ½ Î F ‚Î F q·iÍ € H C Æ Î FÍ 2 ç 5 8·i'@v9æ Ñ ¿÷ˆü Í Ð ½ Î ‚ —E C Ê šÑ ¿£œfbE Ð ½ Î FÍ ‚ C Î ÐÍ v ç H“ …œQŽa ‚ a ç F Ò Ê 5 ‡ ‰Î FÍ 2 ç Ž˜oò•‘v·i'Pò5 ‚ ‚ . Let “ €ga Ð … kˆ f  Therefore. Recall that depends on through .2) (6.

first compute : provided that for (7.1 (Jointly normal random variables) Given parameters: have the joint density Î Î Ó s£êÍ F Í ‚ Ê Î ÓÍ ã â Î ü Í ‚Þ fd'£êþF )‹ÿ÷Ãï Ð and then replace the dummy variable Æ ” ˆ ~f vÎ Í ‚ „ XC ƒ ½ Î ‹  t t ÷¿’â t Í Ó ‚Þ Ð Ê Î ‹ Í Ãï c÷¿8f‚ The function Ï Î F Í ‚ Ê ã â Î ü Í ‚Þ 4fd€F )‹ÿ÷Ãï Ð Ê ‹ ½ Î ‹ ™'d6Í m Vj š › ™ – ”˜ — ˆ  'd’â Í ˆ •f vÎ Í ‚ „ XG%C ƒ C €  t t ½ Î ‹ t ‚ a ο‹8Í ˆ  kˆ f  tÍ Ê ¿’â Î ‹ ” ˆ ~f  tÍ Î F œÆ ü ÿÍ has a joint density kˆ f  In conclusion. to determine Notation 11. Let Ï ´ ¬ ³ µ|¡ ± ¯ ® ²¨ ° i)­ ¬ « xq¡ © ª¨ Ï ” ˆ •f vÎ Í ‚ ƒ„ €EíF Vˆü Í Ãï Ð Î ‹  t C Ê ã ‹ Ê â Î ‚Þ t ½ ¿’â t Í Æ ã ‹ Ê â Î DTcF Vˆü Í Ï f ƒ C € ‹ ™½ t ½ ¿‹ Æ t Í kˆ vÎ t Í ‚ „ XG%C Î  Æ Î Ð 8H“ pÍ Ï Î ¿‹ Æ Æ Î F 82iÍ Ï ‚ Ê ã â Î ü Í ‚Þ d€¬F )‹ÿ÷aï Ð ‚ Ê Î ‹ '¿8Í ” ˆ ~f vÎ Í ‚ ƒ„ C ½ Î ‹  t t ÷¿’â t Í ‚ : “ €~“ #@‚ Ð r Ð p ” ˆ ~f  Ð p Î ‹ r “ #Bd’â t Í A function any function 11.33 If Proof: Just verify that defined by (7.) by the random variable .7 Conditional Density CHAPTER 11.2) (7.§ ¤ ¹ ¢ œ ¢ œ n™œ ¦œ  œ ¬ ½i¥ ¾ ¢  § ¤ ¼ » º Gi¥ (™¸¦œ ¶ · ¿‘¡ ¢ § ¤ ¢ ¨ ¢ ¨ ¶ ¤ %vp¢ ³ ¥ ¥ ¶  ¢ ³ ¢ ± ±   ¥ ¦ § ¦ ¥ ¤¡   Ÿ ¢ œ¡   Ÿ  —•£nX•£"|X•nžœ . then F ‹ (a function of ).1) becomes is often written as does not.4): For v Qç Example 11.1) satisfies (6.1 Let be the function satisfying ‚ and (7. but ” ˆ ~f  ‚ Theorem 7.1) 127 Ê Î ‹ '¿8Í ‚ ‚Þ Ãï Ð ã â Î ü ©F )‹ÿÍ ‚ ‚ . General Random Variables is called a conditional density for given F ü Î • tr Æ »Þ and (Here is the function satisfying depends on .

Therefore. ¢ ¢œ  %§ œ Æ Å PC ‰)•« ´ ä ³ ´ ß ã ‘© ~ áâ ß ± Þ ˆ²¨ ° g® ­ In the -variable. and variance ± § ¤  ¢ Gi¥ ¹ ¦œ ¶ Ô · ¥ ¥ ´ ¬ ³ o¸¡ ¬ o³ ° ¨ )­ ± ¯ ® ²¨ ° i)­ ´ ¬ ào³ ß ± Þ 9i¨ ° e® ­ Ç .  q¬ Qi¥ ¢ œ ¢ § ¤ ¨ ¿ ¢ ´ ¬ o³ ± ß ± Þ ˆ²¨ ° e® ­ ± Ð ³ Ë Í Í Î qÉ Ç ¬ ³ É È { ³ Ë Í ¤ z ÎË  œ Ì Ë Ê Æ ª¢ ¤ Í ± Î qÉ ¢ ™œ Ç ¥ Ç ¡ ËË y¢ Ý ÑÍ ¤ Ñ Ë Ë  ¡ ÑÍ ËË y¢ ¤ Ë  · ¶ Ô ´ ¬ i³ ¨ ° ­ ´ ¬ ³ µo¸¡ ± ¯ ® ²¨ ° i)­ « Ë Ö ™ÎË œ ¸Ë (iÎ 6Ë Û Ú Ù Ø × Æ Î ÏÜÜ Ù Æ Î Î § ¤ ¢ Qi¥ ¹ ¢ œ  œ ¶ · ¥ ¢ ¢ œ Ó %Ð Ç É { ³ Ë Í ¤ z Î œ Ë "Æ   Í ± Î ÏÉ ¿ ÑÍ ËË ‰¢ ¤ ± ÍÌË É È { ³ Ë Í ¤ z ÎË V8|ÊÆ ª¢ ¤  Í Ë  ± Ð B¢ Î ÏÉ Ô ´ Ç Ç PÅ Ó Æ ¢ œ ¿ Å · ¢ ~¤ PC ¥ ¶ Ò ËË Ñ Í ¢ ¤ Ë Ë  ÇŽÅ Ó %Ð § ¤ Gi¥ ¹ ™’¦œ ¢ œ Æ · ¶ Å PC ¢ ¥ using the substitution ±  œ ¬ Gi¥ § ¤ – ¢ ¢ ¨ ¶ ¤ ¥ ¥ ´ ±  ¦œ ± ¢ œ ˜ ³ ¢ œ ¤ ¿ ¢ ¢  ¤  œ "%§ ¢ ³ ¢ ¢ œ ¾ ¢ ˜ ™œ ½¤ à i¬ ¢ §½¤i¥ ¢ § ³ ¨ ¢ ³ ¢ ¢ ¢ ¢ œ ¾ ™œ ³  ¢ ³ "œ Á¬ ¢ Gi¥ § ¤ – ´ ¨ ¶ ¤ ¥  œ ¬ Ài¥ § ¤ ¤ ¢ ¨ ¢ ¨ ¶ ¤ ¥ § ¶ ¢ ± The exponent is We can compute the Marginal density of 128 as follows . Éq Î œ Ë ÊÕ Æ  Ô Ö ´ Conditional density.´ ´ © – Á áâ ± Ð ¬ ³ ß ± Þ 9i¨ ° g® ­ ¢ à ³ ´ ç敫 ¿  œ § ¤ ¢ Ϭ ¢ ½Å¥ ¨ Æ ¢ ˜ ™œ ¤ – PC Å ³ " %§ œ ±    ¢ ¢ ˜ ³ ™œ ¤  `§ œ Å å X³ . From the expressions Ç ¢ ¢œ ¥ ¿ Ë ž¢ Ë ÑÍ ¤ Ë  · ¶ Ô ´ ´ ´ ¬ Äo³ ° ¨ )­ « we have Thus is normal with mean 0 and variance is a normal density with mean .

CHAPTER 11. the best estimator of is .3) & © áâ & Î Î « Ë Í Í Î ÏÉ ú ú tÆ © y Æ – © Á – Ï Ï V)Ï Ï Ï ))Ï Á áâ © Æ )r )Ô Í tÆ t Æ Ô  Ϭ —ˆ¥ ¢ œ ¢ § ¤ ¨ áâ yÍ ióR Ê ù « é (7.4) (7. the th element of is if and only if is diagonal.1).6) are independent . . This estimator is unbiased (has expected error zero) and the expected square error is .3) and (7.. General Random Variables From the above two formulas we have the formulas 129 Taking expectations in (7. has a multivariate normal distribution if and only if the random variables have the joint density Here. and the corresponding Let denote the column vector of random variables column vector of values .e.5) (7. 11. No other estimator based on can have a smaller expected square error (Homework problem 2. .4) yields Based on .e. is the covariance matrix ê Ï û Î ù áTË & Î & ú Î ú ü Ð †ï ¦Æ ¿ ¢ q¬ ¢ Gi¥ ´ à « œ § ¤ ¨ ü PÆ ¢ ˜ « ¢œ ¤ ¿ œ § ¤ ´ ¢ q¬ ¢ Gi¥ ¨ Âà  %§ œ ¢ é 8Í Ï Ï )VÏ Ï Ï ))Ï Ï ñ Ï Æ Æ Ô ü ü r sÆ Ô ÿÍ úÎ ù Ë é |Tó8Í r Ë « ¡   ´ X•« ü ÐÍ œï …óR Ê ¡ è« ¢ ™œ  `§ œ     Ô ´ ä Ên« ¢ œ  œ áâ ¦`§ ¢ é Ð ‘ï cÊ ˜ « ¢ œ ¤  œ "%§ ß ã @© ~ áâ ø ~! é ´ ÷ öõ ôÎ ó ò (ï vXvf‰8Í ñ ð ï ì Ê Î êÍ ë ˆ("î í'÷i4o Ô €ô (7. The random variables in Æ ý µ& Ï Î ù Ë é Î ù TŽ8Í TË ¥ ¥ éÍü Ð 8…gï cÊ î Ê 5 Ô õô Ô õô 5 5 Ô õô 5 y üÍ Ð r Î ¥ PË ¥ ÿ%ï cÊ r ¥ 2 Ô õô º  îPº 5 Î ÿÆþ qs8Í 5 and is an nonsingular matrix. Æ ü where is the variance of . é Æ 8Î r ú 2 )Æ Ï Ï V)Ï Æ Î r 2 Æ Ô 8 W 2Í r r y ~Ë ¥ ü )Î Í ¥¤£   y Ë  Í ¢~¦¡ÿü%ï Ð i. i.8 Multivariate Normal Distribution Please see Oksendal Appendix A.

then they are jointly normal.10 MGF of jointly normal random variables  ! r 2Ô 2 ó ÷ r Ë (ï "¦ ©À ¹ ‰võµXÛ À Ê Î 'ir fk Æ t 8Ô t Í Œ  f  and we have the formula from Example 11.1.9 Bivariate normal distribution 130 in the above definitions. and let have a and mean vector . adjusted to account for the possibly non-zero expectations: Æ r Ô r Ë "¦ À ¹ 92 |2 À Ê 5 ð ï ("î ì  m Œ Æ m PÔ j Œ © ô Œ Œ ô sÔ  j Ô Œ ©  © à r r 2 r 2Ô µvf2 Ë r Ô ˆ2 f2 m ô sÔ j Œ © © Ë Œ m   sÔô Ô j Œ © §¨ Œ Ô Ê 5 Ï r Ô ˆ2 ÷2 Î r y Ë r üÍ Î Ô y Ë Ô üÍ 8že£½ÿ)i÷P|jÿ%ï Ð RÊ ¦ Take Û Ê c%º Thus. Then the moment multivariate normal distribution with covariance matrix generating function is given by If any random variables have this moment generating function. and let . and we can read out the means and covariances.Ï 5 24 Ä iÆ ã Ô fû ÷ (ï Ê y •Á Ä 2 r Þ h ¥ Ä r¥ r¥ Ô ú ¥ ¥ Ï Ï )VÏ Æ Ô ÄÍ )ž…óÊ ) 01 #! 5 4 3¥ 2 ü ¥ Ä Ô fû ú h¥ ) 01 #! ÷ (ï ï cR Ê Ð ë (&$! '%"ï Ð & Î ú Ï à Πr y rr 2 Ë r 8µTr t Í Á Let denote a column vector with components in . The random variables are jointly normal and independent if and only if for any real column vector ú ü PÆ Ï Ï ))Ï Æ r ü Æ Ô 8GP)%ü º Ï Ê Ê ú t ½ Ï Ï ))Ï Ô ½ t 'Î ú tÆ Ï Ï V)Ï Æ Vr Æ t RÔ t Í û ù #Á & Ï))Ï Ï ú ü PÆ Æ)r½üsÆ8Ô%ü  Ï é & 5 ! ÷ (ï r 5 Ô Ô õô Y ô & )VÏ ø Y ! ô Ï Ï Y C Y C é Ï & ! "ï Ð Ê ù é “Ð Ô õô & Î ú Ä ¦Æ Ï Ï ))Ï Æ r Ä Æ Ô Ä )yi8õ…Í Î r y Ë VµP£r r 2Ô µv÷2 Í Î Ô y t )i÷TË Ô t Í ¦Û Ô Á Î "¦ ©D³Û r Ë ÀÍ Ë r 2 Ë Î Ô y Ë À r i÷T'Ô t Í Æ 11. ¦ r 2 ¦ Á Ê Ô €ô 5 11.

As before we denote the column vector P by therefore have for any real colum vector .1 Discrete-time Brownian Motion Semi-Continuous Models Chapter 12 .34 û Ö ¾ a ¦¾ú T¬`Ò ¥ ¥ Proof: is a martingale (under I ). P 131 ã ¾ Ç â  ¦¾ Ô à aÞ aï Ð . Î s Æ Ç Æ £ðPH Í 8 76 & Î ú F Æ Ï Ï )VÏ Æ Ô F ) Í & Æ Ô ÄÍ Ê )ž…óî If we know then we know Define the discrete-time Brownian motion (See Fig.1): 12. Conversely. 12. Define the filtration ¾ ¬a Æ Ï Ï )VÏ Æ r a Æ Ô Vx))Ha Æ Ö Æ d ÷ Ò Ê Ç Ù Ï Ï )VÏ Ôõô8¾¬aT¬ØDØõvÆ Ë ¾ a Ê ¾ F Æ Ô a Ë r a Ê r F Æ Ô a Ê Ô )~棐9)H F Ï Ï V)Ï Ï Ï V)Ï ¾ ¬a Æ Æ r Æ Ô V£a ¨~a ¾ F õvÆ Æ r F Æ Ô )ˆv) F Theorem 1. . if we know . standard normal random variables defined on where I is the market measure. We .Ï ¾ Øa Ê Ê Ê ã ¾ Çâ ¾ ¾ a ØdÁ  ¦€F ï Ð Ô à ¾ a dÁ  ¦¾ F aï Ð Ô à Þ Ù Ï º iÆ Ï Ï V)Ï Æ À Ê oò·Å m)¬a Æ Æ Î ¾ Ï Ï V)Ï Æ r Æ Ô aÍ 2 Ê Î ¾ F )£a )~8£'V€Æ Ï Ï ))Ï Æ r F Æ Ô FÍ T9Rfi£2 6 B DC Ê ¾ ·Ç Ï º ¦Æ Ï Ï )VÏ Æ À Ê #òÝÅ Æ ¥ F Ô fû h¥ Ê Ê ¾ Øa Ù ¾ Æ » a Ï 5 24 r¥ Ô Ä r Ô fû ú h¥ ÷ (ï Ê Î ú Ô fû ú ) 01 A! 5 4 ¥ 2 Ä iÆ Ï Ï V)Ï F Ä ¥ h¥ ) 01 A! ÷ (ï ï cí Ð Ê 9&$! @'%"ï Ð Ô fû ú Ö Let be a collection of independent. then we know . F Ò .

alone. 1 Y1 k 2 Y2 Y 3 3 4 Y 4 .1: Discrete-time Brownian motion. 2 t 0 B k is a Markov process. .35 Proof: Note that which is a function of ¾ ¬a Given parameters: The stock price process is then given by » aä Ù » aä “ î‹y Ð ç t t a `Ò Note that ý 12.Æ Ï Ï Ï )VÏ û Î 2 r r Ô Ë Æ » Ê g·Å Æ yÍ Á Ô à ¾ F ˜|f¦€y2 ÷ (ï ø ! ¾ ý Ê Ô à ¾  cáf¦ ý º iÆ Å û sÎ 2 r r Ô Ë yÍ Á ¾ a %icص2 ÷ (ï ø ! Ù ý Ê ¾ g© ý Æ Î ¾ a 8VØ8Í ‚ Ê ã â Î ¾ a d€R¾ Ç )TØdÁ f¦€i÷Ãï Ð Ô à ¾ F Í ‚Þ Y ô Ï ‹™½ › r H8GØ8÷‚ !Î E Á ‹Í ô Œ Ô Ï Y C ó XÛ À ì Ê Î ÷8 E FÁ Ô à ¾ FÍ ‚ Ð Ê Î Í f¦€i÷‘ï c÷8 f‚ 7E ã ¾ â Î ¾ a ¦4Ç )TØdÁ  ¦€Í Ô à ¾ F ‚Þ Ð Ê ã ¾ â Î Ô à ¾ a Í ‚Þ Ãï c€R·Ç )ifi¬8÷Ãï Ð û ¦¾ú Ö ¾ Ù Theorem 1. Figure 12. the volatility.2 The Stock Price Process Then Use the Independence Lemma. the initial stock price. Define 132 . . the mean rate of return.

be a probability measure on Î P¬ Í Ç Æ 76 is a martingale under ÑÐ . we say that ÑÐ Y ÑÐ Y Y Ù û ¦¾ ú is a risk-neutral measure. is -measurable.Ï ¾ ¾ ðü ñ Á ˜ ¾ ñ ¾ fý Ë Ô à  ¦¾ Ô à ¾ f¦ ý ñ Ô à – ·gÊ fi¾ ñ ¾ ù Ô à ¾ fiÃü Ê Ê Ô à ¾  ¦ðü Î ¾ ý ¾ ù Ë ¾ ü V )Ý´©ÃÿÍ ¾ðü U !GÁ )¾fý U Xáfif…d·ù Î ! Ë Ô à¾ ýÍ ¾ ! Á Ô à¾ ý¾ U W|f¦ )·ù Æ Ô )õô ú Ï º ¦Æ Ï Ï V)Ï Æ À Æ » Ê Å #)g4'Æ ¾ U! Ê VT¬¾ ñ Ï r 2 Ê { ·|Î 2 r r    Æ Ô ¾ Ç Ë yÍ %ivÁ fi¾ Ô à ¾  ý   Ô à ¾ fifý ¨ ©r ô Œ Ô 133 P ã¾ %! Ï 82nj â Ï ¾ ý P ! !¾ © Ôr ô P ! Œ 2 Ôr Q ý I!Ãï V ý Þ Ð ¾ Ê Ê Ê Ô à ¾ F f¦€y2 The other processes in the market are defined as follows.1 Let ù iÆ Ï Ï V)Ï Æ Ô ù 8Ã¦Æ Ù Money market process: Discounted wealth process: 12.4 Risk-Neutral Measure Portfolio process: 12. equivalent to the market measure I . F y2 z ï XV7Ê Ð W US ¾  ý ã R¾ Ç mf¦ tÃï Ð â Ô à ¾ ýÞ W US Ê XV7îy Thus ã ¾ â Ô à ¾ ýÞ m·Ç )fi ›ðï Ð and S '¤ R Ê ˜ ¾  ý Ô à ¾ fifý W US XVX– S @¤ R Definition 12. . -measurable. nonrandom. Semi-Continuous Models Each is ¾ ·Ç Wealth process: ¾ Ýù Each given. If P û ¾¾ ýñ ø ¾ ·Ç ¾ Ãü t Ù ü t t ù t t .3 Remainder of the Market CHAPTER 12.

Hedging a short position: Sell the simple European derivative security Ù ü . 12. û ¦¾ Y Theorem 4.6 Arbitrage ú ü » cÊ ü Definition 12. Ï » ä Î » €#aä Ï ú Ù ú ñ ¤ ïÐ Y Ê ú ú üñ ïÐ Y Ê Ù ü ÑÐ Y If there is a risk-neutral measure .1 Hedging in this “semi-continuous” model is usually not possible because there are not enough trading dates. then every discounted wealth process ú û ¾ Ãü ñ ¾ ø is Ù ¤ t t t t . then there is no arbitrage. This difficulty will disappear when we go to the fully continuous model.5 Risk-Neutral Pricing ú ú Ç º Let be the payoff at time . Ô õô ú ù iÆ Ï Ï V)Ï Æ ù Remark 12. regardless of the portfolio process used to generate it.36 If is a risk-neutral measure. . and say it is ú -measurable.134 Ù ú ÑÐ ÑÐ Proof: 12. Receive at time 0. then and ends with satisfying Ê ú ü ü Construct a portfolio process which starts with and ends with ¤  ¾ ¾ ðü ñ ¾ Ç     ¾  ¾ ðü ñ    ¾ Ô à f¦¾ Á ˜ Ë ·Ç ¾ ¨ gï ÐY ~Ýù – ¾ ¾ ý ñ Ô à ¾ f¦ ý ñ  Ô à ˜ ¾ ñ Ë f¦¾ ñ ~·ÿ¨ ï ÐY – ¾ ù ¾ ý Ô à ¾ f¦ ý ¾ ¾ Ù ¤ üñ Á ú ¤ Ï ` aú Ê Ê Ê üÍ ÿ·Ñ Ð  ¾ ·Ç     Ô à f¦¾ Ô à ¾ fiÃü ñ ¨ gï YÐ Y a martingale under . P Theorem 6.37 (Fundamental Theorem of Asset Pricing: Easy part) If there is a risk-neutral measure. Note that may be path-dependent.2 An arbitrage is a portfolio which starts with ú üÍ Ð Æ À Ê Î ÿ·Ñ ÷#ò'#» (I here is the market measure).

. 12. If . CHAPTER 12. the market measure is risk neutral. We have . and let be the final wealth corresponding Proof: Let be a risk-neutral measure. Since is a martingale under .1) 135 .2) (6.7 Stalking the Risk-Neutral Measure This is not an arbitrage.Ö r 2 r Ô Ë Â ¬¸Ë y oÒ ! ÷ ï (vÏ Ö Ö r 2 r Ô Ë Â Ë y ¬‹îÒ ÷ ï (Ï ! r 2 r Ò Ô Ï Ï ¾ ¾fý ñ ô ÷(ïvÏ ¾ U P! Ê ¾ ýñ ¾  ý ñ Ê ã ¾ â Ö Ô à ¾ F 2 8·Ç áVfiõ‰%Ò ÷ (ï ! Ê     ¾ ðÇ ! F ˆ2 Þ •ï Ð Ï ¾  fi¾ Ô à Ô à ¾ fifý ñ ¨ ïÐ Æ û Î 2 r r Ô Ë Â Ë yÍ ¬‹îicÁ fi¾ Ô à ø ! ÷ ï (vÏ ¾ ¾ fý ñ Ê Ô à f¦¾ Ô à f¦¾ ý ñ Ê Ê Ô à ¾ fifý Î À Á ÅÍ û )H…VÎ 2 r r Ô ÷ (ï Ï ûÎ 2 r Ë ˜|f¦€y2 yÍ Á Ô à ¾ F ¾  ý r Ô ø ~! ÷ (ï yÍ Á Î Ô à ¾ F Á ¾ aÍ ˜õRf¦€dØ6£2 ý Ù ø ! Ë Å û sÎ r 2 r Ô Ë yÍ %icÁ ¾ a µ2 ú ÷ (ï ø ! F vÆ Ù ý cÊ ¾ ý Î s Æ Ç Æ £v%s¬ Í Ï Ï V)Ï Æ r F Æ Ô )ˆv) F 76 t t t Ï » Ê Î c£o» ä ú ü ·Ñ Ð Í ‡ Ê » Ê Î » 9H#aä ú üÍ ÿÃÑ Ð Y ‡ Ê À Ê Î » ˆ©ò'#cÊ À Ê Î » ò'#cÊ ú ú ü ðÑ Ð Í Y Ï À Ê Î '#» ` ¡ú üÍ ÿÃÑ Ð Y ‡ Ê » Ê Î » ˆ¬c'#a¼ ú ü ðÑ Ð Í Y ‡ Ê » Ê Î » ˆ¬c'#a¼ ú üÍ ÿ·Ñ Ð Ï » cÊ h ú h fb Y iTï Ð#Ê ú ü ñ ï ÐY Ù û ¦¾ ú ÑÐ Y . let to any portfolio process.1) and (6. standard normal random variables on some probability space . we must seek further.2) imply üÍ ÿðÑ Ð Y If Therefore.  gÊ d Wy Â Ê cîy are independent. (6. We have . ú û ¾ ü » ˜Ê Ù ü ¾ üñ ø ÑÐ Y Suppose ` Í aú ÿü·Ñ Ð ‡9ÊÀòÊ'Î#» ` Í cú ÿü·Ñ Ð À Ê Î ò'#» (6. Semi-Continuous Models Recall that .

standard normal ran- ` Î 5Í Ø8ÃÑ Ð Y t t . and are independent. for all Ç ç j˜5 » In other words. Then we would have Æ Ô ¨fF Y U ô© P ø ! ø ! ø x! ÷ ï (vÏ ¾ È ÷ ï (vÏ ¾ ÷ ï (vÏ ¾ ¾ fý ñ ¾ fý ñ ¾ fý ñ Ê Ê Ê Ô à fi¾ Ô à ¾ f¦ ý ñ The quantity where 136 is denoted and is called the market price of risk. ÑÐ Y is a probability measure.Ï À òÊ é ï g'© ðÑ Ð Ð Ê Î Í 76 Y Ï Ç ç 5 j˜9æ Ñ ¿½ Ð é E C Ê Î 5Í %'¬8ðÑ ÐY Û º Ï Û º À òÊ Ë ¡ r g Û º Ë ÷ ï (Ï   ! F g Ë %hDÍ ¡ Î r ú g h ) 01 ¥ ÷ (ï   ! ! ÷ (ï ïÐ Ê Ê ¡ r g ÷ ï (vÏ   ! 5 24 Ô fû ¥ ïÐ é » ` é Ï Î r Ô gr Ë ¥ F g Ë ohDÍ Ô  û ú h¥ ÷ (ï ! Ê é  §¨ é Ï ¾ Ö Ö ¾ ýñ Ê Ê r Ë 2 r dÒ Ô ÷ ï (Ï ! r ý ÷ ï (vÏ ¾ ÷ ï (vÏ 2 Ôr dÒ Ë Ö 2 r Ò ¾  ý ñ r Ô ! ! ¾ Ç â Ö  ¦¾ Ô à ú    Ê F `Ò 2 ! FÆ È )VÏ Ï Ï ÷ (ï ü ïÐ YÏ ¾ ¾  ý ñ  ÑÐ ¾ Ç  Ô à  ¦¾ Ô à fi¾ ý ñ ¨ ïÐ Y Ï È È P Á Ô à¾ F Ê Ô à¾ U ô© f|fiõ7áf¦€F Ë Ô à ¾ fiõF Æ û r 2 r Ô 2 û r r û (Î 2 r Ô Ë 'Î P Á Ô à¾ Í U ô© e' ¦€FÈ |2 2 r Ô Ë Â Ë yÍ Á Ô à ¾ F ¬‹%i˜'fiõ‰2 under which We want a probability measure dom variables. g Then Define Properties of : Cameron-Martin-Girsanov’s Idea: Define the random variable .

Semi-Continuous Models Verification: Compute the moment generating function of Ï g WÁ ú F 7Ê Ô fû ú  F Ä ¥ ¥ h¥ ÷ (ï §¨ ! ïÐ almost surely. is a risk-neutral measure. and P for every random variable .Ï  Î { |Î Î r Ô g r Ë H g ¥ ÄÍ Á …c Î Ô  û ú r Ô g r  g ¥ r Ä ´Ë Ä Î # r¥ Ô r ¨z Í Ä pÍ Ä r Ô fû ú Ô fû ú h¥ ÷ (ï §¨ §¨ §¨ ! ! ÷ (ï ÷ (ï ! ! ÷ (ï Ê Ê Á © r¥ Ô h¥ h¥   r Ô gr Ë Ø Ë ¬ g ¥ Ä …Í Ä …Í Ô fû ú h¥ ÷ ï (vÏ §¨ ! Ô  û ú ! ÷ ï (vÏ   ¥ g XË ¥ ¥ F iÍ ¥ ¥ Æ Ô Ô r Ô fû ú ÷ (ï Ê r Ô Î gr  g g ¥ h¥ r Ô r Ë §¨ FÎ g |#XË Ä …Í Ô fû ú Ô fû h¥ ïÐ Ê Ô fû ú §¨ §¨ ¥ F g Ë ohDÍ ÑÐ h ¥  TFÁ Á Î g Î ú Ä ú h¥ ! ïÐ Ê  F Ä ¥ ¥ È h¥ ÷ (ï §¨ Î ! ïÐ Y È Æ Ç ç 5 %%v9æ È Ñ ¿½ Ð é Ï  r¥ Ô Ä r Ô fû ú h¥ ÷ (ï §¨ ! Ê g FÁ ú F 7Ê ú F Æ È Ï Ï V)Ï Æ  g WÁ Ô are independent. it suffices to show that : Independent. standard normal under ÑÐ Y F 7Ê Ô F We show that È ÑÐ Y CHAPTER 12. standard normal under I . . under Y FÆ Ï Ï ))Ï FÍ : 137 ü üÍ Ð Ê ü ÿ%ï cyHï Ð é Y t E C Ê Î 5Í '¬8ÃÑ Ð Y ý r Ô Î Æ g r Ë ¥ ú F g Ë Ô %hDÍ fû ú ¥ Ï Ï V)Ï piü ! Æ  ÷ (ï Ê » €ä Ô F f7Ê é t éF F Æ È g WÁ t t È ú F vÆ Ï Ï V)Ï Æ r F Æ Ô )ˆv) F t . For this.

E‚½ s € Œ Œ ô Œ ! º ó  4Û À ì Ï à { º sÎ q Ë ûº X·isÎ r r 2 r Ô 2 r Ô Ë ÂÍ ©…vÁ E y2 ¥ ÷ (ï ÷ (ï ¥ ø b! ! È Ù Ù ý z ú ý r i Uô! ô vxwquË 24 5  à Ë ÂÍ ¬…šÁ ¥ È F Ô fû ú h¥ 2 ) 01 rt sú U ! §¨ à Ï 5 24 º PÎ r 2 r Ô Ë Â ©pÍ Á ¥ ¥ ¥ F F iÍ F Ô fû ú h¥ 2 h¥ 2 ÷ (ï Ù ý ) 01 A! ) 01 ! Ê 5 24 º sÎ r 2 r Ô Ë yÍ Á º Î Â Ë yÍ Ë %i˜†sT´%ia'Î Ô È fû ú Ô fû ú ú ÷ (ï Ù ý P U ô© eÁ Ë yÍ %icÁ Ë Ê 5 24 º sÎ r 2 r Ô h¥ 2 a ˆ2 ÷ (ï Ù ý ý Ù ) 01 A! ø ! Ê Ê º ú ý º û PÎ 2 Ôr r yÍ ˜Á ÷ (ï Stock price at time Payoff at time 12. under ÑÐ Y º This is the Black-Scholes price. variance . Y Y ô C Ê ïÐ ú Y Ê à Î q XË ñ ú º ý …Í ïÐ Y Î q XË ú ý …Í .8 Pricing a European Call 138 is is since . Ô F fû ú y . Price at time zero is is normal with mean 0. It does not depend on .

2 The Law of Large Numbers We will use the method of moment generating functions to derive the Law of Large Numbers: Theorem 2. as Æ Ï Õ cÊ Ú Ê Ï À ¥ ` Ó Ó Å if if ¥ Æ ¥ ü À HË Ô fû ¾ À  » cÊ h¥ Ê ¾  Ê Î Ó '£êÍ ñ » r a¾ ¥ ü ñ À Å .1 Symmetric Random Walk Toss a fair coin infinitely many times.Chapter 13 Brownian Motion 13.38 (Law of Large Numbers:) 139 Ï  r d"Å almost surely. Define Set Ù ñ 13.

Æ ¾ { ‡ … “ ô ! Ôr ¡ ¾ ñ Å ì Ä Á ‡ … “ ! Ôr z ÷ (ï   ! Ê ï c'³…¿¾ Ð Ê Î ÄÍ ƒ r ¾ ñ Å À Ï ì  ‘ ! Ôr r Ù r S   ‘ ! ‘ ¥ˆ£ z W U ½XVS Ù {  ‘ ô ! rÔr Á ‘ô!‘ Ë t r ‘ r  ‘ ô ! Ô Á  ’! Ô { X‡ … ô ! Ôr ¾ Æ ü { 4‡ … ô Å ü Ä ¥ ¡ ¥ ¥ Ï ¾ ñ 5 24 ü Ä Ô Å fû ¾ ¡ ¾ Theorem 3. of  r d"Å ! ÊÎ Í T'³…Ä¿¾ Y r ¾ ¥ˆ£ S ¾ Ê Proof: (Independence of the (L’Hˆ pital’s Rule) o ) ’s) . ƒ » cÊ Ê Ê Î ÄÍ '³…¿¾ ¥ˆ£ S r  W U XVS Y r¾ ƒ ƒ ‰ˆ£ S z W U Å Ê Î ÄÍ —4fS c'³…¿¾ W U XVS which implies. as (Def. 13.g. Then Standard normal.f. Ô t Let 140 . Á ¬ ‡†! Ôr … z ïÐ Ô fû ¾ Ê „ ¥ Ê ÷ (ï ) 01 !   ! ï cÊ Ð ï c'³…Í ¾ Ð Ê Î Ä ÷ (ï Proof: ƒ Therefore. for the constant 0.39 (Central Limit Theorem) We use the method of moment generating functions to prove the Central Limit Theorem.3 Central Limit Theorem Æ À #òÊ Ù Ï Á ¬ ‡! Ôr … ! Ôr ÷ (ï   ! h¥ Å ñ Ä which is the m.

4 Brownian Motion as a Limit of Random Walks r ‘ô! ‘ (L’Hˆ pital’s Rule) o Ï { ‡ … “ ô ! Ôr Á which is the m. 13. Therefore. then define . If ` • º ¥ˆ£ S If is not of the form .1). then set : ` • by linear interpolation (See Fig. (L’Hˆ pital’s Rule) o 141 . Æ Œ Ô ! ÊÎ Í ‘ r T'³…Ä¿¾ ƒ Y r ¾ Û Ë r ‘ô! ‘ Œ r t Û ‘ r Ù S  V! ‘ ‰ˆ£ Œ r Ù Ù Ï Ä r Ê r Ô r  Ê Ê ‰ˆ£ S r r r   Ê t Û Ë ‰ˆ£ S ‰ˆ£ S  Ù Ù Ê Ê Î ÄÍ 'r…d¾ ‘ô! ‘ Ë ‘!‘ r Ï ‘ ! r Á  ‘ ! Ôr S  ô Ô r Ù À  ‘ ! ‘ ¥ˆ£ {  ‘ ô ! Ôr Á  ‘ ! Ôr z t Û r r ‘ô!‘ Ë ‘!‘ r t { ‘ ô ! r Á ‘ ! r ‡XVS  Ô  Ôz WU ‰ˆ£ S r  W U XVS ƒ ¾ Ô ‡ … “ ! Ôr z W U Å Ê Î ÄÍ —4fS c'³…¿¾ W U XVS ƒ so that. Brownian Motion Let be a positive integer. Then Î yÍ Ù sÙ Ô j a Here are some properties of ú » Î DÍ is of the form .g. for a standard normal random variable.–• m a » –• m ú j a ì À ¾ Ï ¾ º ñ À ì Ê ú — ˜ñ ú ¾ º Ê Î 'D Í –• m ú j 13.f. Y r ¾ ¥ˆ£ S ” Ê t Let CHAPTER 13.

) 13. . .1: Linear Interpolation to define Î yÍ Also note that: Properties of Properties of is a continuous function of . Chapter 2. let k/n in a (Please refer to Oksendal. ` •dÆmy•–Í ú Î m j • » m Ù sÙ Ô a Ë Î Û j T'o…Í m Ù sÙ Ô j a and Î À )þÍ To get Brownian motion. (Approximately normal) (Approximately normal) Î À VDÍ ÎD–Í m • Ôj a ÙsÙ j a t m Ù sÙ Ô t (k+1)/n .Ï Ï Ô fû ü Ù sÙ Ô  û Ù sÙ » )À À ÛcÊ÷ÎsÎ#pÍ m Û » Ê Î Û c÷#…Í m m Ù sÙ ÙsÙ Ô Ô Ù sÙ Ù sÙ Ô a 8Í Ôj S '¤ a j ‘ï Ð R ¥ r h¥ Ê Î Û ÷#…Í Ù sÙ j a Ô Ô p Î Û Bo…Í m » » #)À À Ê÷ÎsÎ)þÍ m À Ê Î À ÷)DÍ j a a j 8Í À Ê ò£Î ¥ ü ÿÍ Ï S'¤ R ¥ ü †ï Ð ü h¥ h¥ Ô » cÊ Ô Ô fû Ù sÙ Ô fû Ù sÙ Ô » )À À » )À À S '¤ R m m Ù sÙ a j ‘ï Ð Ô ¥ h¥ Ê Î À ÷)DÍ Ù sÙ j a Ô p Î À BVDÍ m Ù sÙ j a ™• m ú j a 142 Figure 13.5 Brownian Motion  r •ˆº are independent.

2) is called a Brownian Motion if it satisfies the following are independent. Then and are jointly normal. Moreover. be given. is a continuous function of . » Ê Î »Í g'#…ða 2. are independent.6 Covariance of Brownian Motion Ï ÿ ¢æ Ô õô Æ ÿ d¢æ ¥ • • Ê S eË ¥ A'Î ¥ FÍ '¤ ú F vÆ » cÊ Ï Ï V)Ï Æ r F Æ Ô )ˆv) F ¥ F ï ÐR t t t –• Æ 8Î Ù –•ÃT'RP–•ð F Í a Ë Î Ô Í a Ê Ô • ¼ ԕ • Ê WØPW¼ A†» Ù Î Í y Ãa –• CHAPTER 13. • Î Í D ða –• 3. Brownian Motion (Ω. Î Í ) Ãa Figure 13. 13.ω) and Ê Î Í îD ða –• Î Í ) Ãa 7f Ï f jÊ i – ™ Ù ™f Î)hÍ ˜)— a‘ï ‹Á sV7ð'y™ð8)V7ð‘ï gÊ Ð Î Î f Í a Ë Î • ˜ Í — a Í Î f Í a Ж r ãB)7ÃadÁ P)7ÃT'y–Ã8ë8V7ð‘ï g£y™ðvV7ð‘ï Ð Î fÍ Î Î fÍ a Ë Î •Í aÍÞ Î fÍ a Ð Ê Î •Í a Î fÍ a Æ • ¦AÊ'sy™Íð8Í Î Î • a Æ » Ê Î •Í a g£y™ð‘ï Ð S@¤ Æ ofT'sÎ)hða8Í Ê Î f Í Æ » Ê Î f Í a S @¤ R R g£)hð‘ï Ð Î Í a Ë Î Í ) ð7Hy Ãa f h –• 7f Î Í a Á Î Î Í a Ë Î Í a ) Ãd sV ð'y Ã8Í Ü € c» Ü 7f 7f –• • f g g 143 t .Æ Î Ô 8)€ô ú –•ÃT'Î ú –•ÃÊ Í a Ë Í a ú F ú Ï Ï ))Ï Æ Î Ô Í a Ë Î r Í a Ê r mRs ðT'8| ð9F Ï Ï V)Ï ™• • W¼ ¼ r ~¸ A random variable properties: 13. a Let then and has independent. (see Fig. so B(t) = B(t.2: Continuous-time Brownian Motion. normally distributed increments: If .P) ω 1. F.

ÎH“ …Í vdç £ Р㠕 »Þ …HÆ t˜ç f Ö £ ÝQçáÓDo7ÍðXp mvØ·QV7ð%Ò Î Æ f a Ó Ò Ê Ö £ ç Î fÍ a ÎH“ …ÍHQçó£ Ð v 㠕 »Þ pHÆt×çœf • Î •Í y–îÇ Î •Í D–îÇ Ï ú –•ÍÃTË'Î ú ™•ða Æ )Ï)Ï mÎ Ô ™•ÍðaT'Î r –•ÍÃa Æ)D–•ðT'Î Ô –•ða a Í Æ Ë Î Í a Ë Í ú •W¼#Q%rT~¸•W¼ØP•W• rr ¼ r Ô ¼ • –• Î •Í y–Ãa • in .2 Î €ô Ô Ù Ö Î Í Ç 7q— ¦D–•î)Ò Ï Ï V)Ï ú r Ô Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï 88V88)88V888)8Ï Ï Ï V)Ï • • • Ê r ¸• r• Ô ¸ps• Ï Ï o  s• V)Ï sps• n¨ Ô Ô • Ô o Î Ô Í a Í a n Ï Ï ú Î –Í • ‘ï Ð V)Ï 8¸–•ÃvÎ ú –•Ãa ï Ioš8s–•ÃvÎ ú ™•ð‘ï Ð §n a Î r Í a Í Ð Ï Ï Ï Ï Ï Ï Ïr Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï Ï 8)8)8V8)V8)8)8V)8V8)8)8V)8V8)8)8V)8V8)8)V8)8V8)Ï Ï Ï ú Î –ÃvÎ r –ð‘ï Ð • Í a • Í a ÏVÏ)Ï Î r –•Í ‘ï Ð Î Ô ™•ðvÎ r –•Ãa ï Ð a Í a Í o  Î ú –•ÃvRP–•ð‘ï Ð V)Ï 8¸–•ÃvirÎ s–•Ã‘ï Ð is–•Í r a ï Ð nn¨ Í a Î Ô Í a Î r Í a Ô Í a Î Ô Ioo §n Î PÎ ú ú Ê v£ –•ð)Æ Í a • W¼ Ï Ï ))Ï Æ 8Î •Í Æ •Í a r –Ãa 8Î Ô –ð8Í • ¼ ԕ ¼ WØPW2» Ï Ï V)Ï ¼ r ~¸ ïÐ » Thus for any 144 . Î Í y îÇ –• is Î Í D îÇ 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 . Do this for all possible numbers required by the -algebra properties. Then 13. (not necessarily Ï •mklfTÊ£ÎD–•Íð)7fÃa aÎ Í • Ü Wœf ` • Required properties: 13. Let . the Brownian motion increments ). -measurable. Then put in every other set be given.7 Finite-Dimensional Distributions of Brownian Motion Let For each . First fix . .8 Filtration generated by a Brownian Motion is jointly normal with covariance matrix » ` f be given. Put the set • Ù Ö Î Í Ç ¦D î)Ò Î Í D îÇ 7q–•— –• t t . we have and and . . is called the filtration generated by the Brownian motion.

‘ “Down” factor. 13.41 Let é “ î2g Ð ç Ï Î Í a V ðDÊ Î Í a V ðdÁ B) ðæ÷D ðÃï gÊ ã Î Í a Ë Î Í aÞ Ð ã Î Í â Î Í a Á Î Î Í a Ë Î Í a ÍÞ Ð Ê ã Î Í â Î Í aÞ y) îÇ )V ðdfs) Ãæ÷y ð8ëÃï gžp) îÇ Vy ÏÃï Ð 7f 7f 7f 7f f h ™• –• 7f 7f –• Theorem 9. 145 . t t t t t ½ Ä ú Þ Ãï Ð . be given.10 The Limit of a Binomial Model is a martingale. Consider the ’th Binomial model with the following parameters: 7f –• g h “Up” factor. Then be given. ” Ê s t s t . Brownian Motion Proof: Let • Ü f Ü WœG·» Proof: Let 13.9 Martingale Property CHAPTER 13.Ô r Ê ú õ 2r õ 2 ú ” Ï ú r Ê ú Ì Ô ú ô s Ê ô sÔ ÈÉ » Ê cØ ú È Ë À òÊ Ê Á £À º       Î Í ) îÇ ¦V Ö Î Ï Î f )7Í Ê ÷ û()fsË –Í g r 'P)7Ãæ'D–ð8Í Î • r Ô Ë Î Î f Í a Ë Î • Í a '¤ r ¨hþÍ Ôr ~! (ï )7fÍ é Ê S Î g Ë ÷ø Î  Î f • r Ë Î Î f Í a Ë Î • Í R a Í g Ë (ï Ð Î f é Ê  Ö¦VXË –Í r g Ô 's)hðæy™ð8hdÒ ! ü ï s)7Í ý  ÷ f Ë X|™•Í r g Ôr 's)hðæy™•ð8ahdÒ ! (ï V7fÍ ©ï cÊ Ë Î Î f Í a Ë Î Í Í g Ë Î ¨ Ðé 7f Î Í ) jÇ ¦) Ö Î f h f Á Îf Ë Í G )X÷–•sÍ r g Ôr Ë Î Î Í a Á Î Í a Ë Î Í aÍ 's) ðd÷) ÃT£D ð8 dÒ Ë f h û • r g Ôr Ë Î Í a y ð% Ë ™• h g ÷ (ï ø ! Ê Î £D Í –• Theorem 9. Then ).40 Brownian motion is a martingale. ( » aä 2 Ï ú 2” 2” ÷ (éï ! ¨ Ð Ê ãÎ f â Ε ©ï c€BV7Í Ç )y–Í • Ü f Ü WœG·» é . Then be given.

steps per unit time. and let u Å In the ’th model. approaches the distribution of 2 r Ë Î Í a £D ðˆ2 –• Î yÍ ý W U á4fS  r •ˆº converges to the distribution of is necessary in order to have a . Note that the correction martingale. Set . • • r 2 r Ô Ë Î Í a 2 £D ðˆ`Ò ý –• ÷ (ï a º Î Å ÕÍ ¾ …dP Under . the distribution of is a martingale. take which implies. the price process . the distribution of Î DÍ !  r d9º ÑÐ Y u As so Proof: Recall that from the Taylor series we have .•r Ô –• m ú j ñ À Î r Ô ô º …Í ™ Ù r ˜ )— º – ñ À w ˜— ú º – Á Î v r ™ ú € )— º ì – ˜ ˜ – 2 PdÁ — º …Í º º ì 2 Ô Ë r Ô ô Ë º }~Î r õ x ô …º%sÁ Í w ˜ º r 2 À | Á º r º 2 À ì r 2 Ô r Ë º Á º ì w Á €• r r º 2 r òÊ Ë Ô ñ Á | 2 Ô z —ú ì 2 Ô Ô }~Î …º%sÁ 2 À | Í w rõx ô r ì Î 2 ˜Î º Ë À Í W U áþžXVS ì Ô r 2 À ì r 2 º | Ë r ú z •º {†cÊ ñ Ô Á Ë Î Ô º Á À Í W U £D€4fS Î º ì 2 Ë À Í W US D€4f¦Î ú 2 Ë º ' …Í Ë À Í W U D XVS r ì ì r – — Á fÎ 2 Á À Í W U £D€XVS ú r – •º †cÊ Ô — ñ Æ 8Î • † Ô r Á õÎ 2 r Ë Á À Í W US 'DõXV"Î — ñ Á º   …Í • † r Ê Î y Í –• m ú j ý W U áXVS x w yt Í WÁ r r 2 Ô Ë r t Ô t Ê 'Î Á À Í W U t £DõXVS Æ Ö ds –• m ú j Ï ¾ m ú jÙ Ï Î )¾ Î ¾ ñ ñ Ë Å 2pÍ Á Å HpÍ Ô Ô r r Ê Î ÚÍ ¾ 'ØÿdP Ê 'Î Õ pÍ ¾ u u Æ ¾ Ê Î Ú 'ØÿÍ ¾ ñ Æ Å Ê Î ÚÍ ¾ õg'ØÿdP Ú Î ÚÍ ¾ Øÿ¿s Õ …Í ¾ ÕÍ ¾ …¿s uu Î Ë u Á F³Î u u Õ Let first 146 denote the number of tosses. Let denote the number of for some . Then in the first tosses. and let Å v m Ds b ô ú j —ú Ô r ˜ Ê ' º ì 2 Ë À – À òÊ • v m Ds b à ú j — Ô ý r ˜ º ì m új ý 2 Á À £– Ê Î £D Í –• m ú j ý º Theorem 10. in the .42 As where is a Brownian motion.

11 Starting at Points Other Than 0 (The remaining sections in this chapter were taught Dec 7.CHAPTER 13. Px) Figure 13. 13. P  Ž ‰Š … Œ‹ ‡ '‚ € ‚ ‡ ƒ a7„…  ‰Š € „ Œ@‹e‚ ‡ ƒT‚ ‡ ƒ …ah… ‰Š  ‰Š ›› ‡ ‡ † ‘ … ˜ œH~sl7’a7„… –š™ ‰— ›› ‡ … ˆ™†a„ ‡ … ˆ™†a„ For a Brownian motion that starts at 0.43 Brownian motion has the Markov property. we have: ƒ ‚ T˜€ .3). Proof: Let be given (See Fig. then puts all its probability on a completely different set from I .  ‰Š “ • “ ƒ –† yƒ ”’ ƒ ‚ T˜€   . starting at 13.ω) x ω t (Ω. Independent of -measurable ›› I°°ˆ%7’… œ‡ ¯± ‡ ›› ‡ … „ ‘ ~–†aW€ ‡© § … ¦ Q7’¨a„ ¥ ­« ­« ®¬¥lª ®¥¬lª ‘ Q7’aX¤¨~¦‡§ W¢7’a@… š Ÿ ‰— ‚  %h… ‡© … „ £ † ‘ … „ ¥ ž  ¡ ‡’ ‡ … ˆ™†a„ The distribution of under is the same as the distribution of under I .4). and for such a Brownian motion we have: Note that: If . Brownian Motion 147 B(t) = B(t.3: Continuous-time Brownian Motion.) For a Brownian motion that starts at . denote the corresponding probability measure by (See Fig. 13. P 13. F.12 Markov Property for Brownian Motion We prove that Theorem 12.

restart s and define s+t . Then 148 Example 13.5. Fix ›Ú ›› Ò ½ÐˆÊ ·–ÍÏ Ð Ï Å ÐÐ Ï Û Ð Ð ÐÊ Ø Ï Í Ö Ü®'Ä ·»ÍÏ eÅ ¾ 'Ä Ï › 'ˆÙlÄ ×%Ï –Ö˜ ÝlÕ Ô ÕÔ ÒÓÀ ˆÊ ·Î@à ¤~¤pÈX·Ä Ñ Å Ð Ï Í Ì Ë ÊÉ Ç Æ Å Ã Á ÂÀ ›› †ˆsl7’m7„… ˜š ‰— ‡ ‡† ‘ … Then we have: B(s) Figure 13. 13.‡ Œ ¿‡ˆ–†…m7„… š ®¥¬´¸ — ‚ ­« ‡‡ … „ HQ7’ah… ² ‚ hQ7’… ¾‡ ›› ‡‡ … Œ Hˆ–†m7„… š  ‰— ‚ ­½‰¬«¢¸h¼@»­ºŽ¹'¥¬´¸ « ° ¯° ± ‡ §a ˆ©‡ ™†¨¦ „ ¥ … ‘ € ·'… š ž  ¡ ‰— ‚ ¶‡ € ‘ ‡ ’… „ £ ‡† ‘ ’… „ ³ ˆµG”%haX¢ˆs´ha'… IšG ‰— ‚ 7€… ² ‡ Use the Independence Lemma.1 (Strong Markov Property) See Fig.4: Markov Property of Brownian Motion. Define same distribution as In fact Brownian motion has the strong Markov property.

5: Strong Markov Property of Brownian Motion. † € ‡ D• € • ™†Þ ß … Fix Fix 13. x restart τ τ+t 149 . and Let be the probability that the Brownian motion changes value from let be defined as in the previous section.14 First Passage Time CHAPTER 13. Brownian Motion 13.þ ‡à ü … „ Œ‹ ‡à 'e‚ ywü –†… ÷ ÿ£ yðý–†aœ÷ ¡         û 3’ø ‰— úù is a martingale. and ‡à yðü ™†… ÷ ÿ¤yðý–†mC÷ þ £‡à ü … „ û úù ·3’ø ö ‡ … õ “ ô ò ñ Œ ¢€ ‚ ˆ™†a„ ƒ –† Ž¥ó"T‚ à ¡ ï ƒ "÷ ï ƒ ð€ ì ¼ ›› ‡ à ›‡‡† ‘ … Œ î ß … ß ‡ D• € • –†7އ ß … š ì ‚ ¾ yh… › H~s”7àm7„… ˜š ‰— í ì ¼ ›› î … ‡‡ … ‡ › ‡‡† ‘ … Œß ‡ ßD• ‡%7’…m„ • ™†7އ ß … š ì ‚ ¿%7’m7„… ² ‚ ¾ Q7’… › Hˆ´7’m7„… ˜š ‰— í ì ¼ ‡‡ … ‡ Œ î ß … ß ‡ D• € • ™†7އ ß … š ì ‚ Hˆ™†a7„… š  ‰— ‚ 7€… ² í ë ä ®ê C† ß … pê~C¿Iæl¼ '‹ â á ‚ ‡ D• € • –†Þ éèç å ã ß to in time . Then .13 Transition Density . Define à Figure 13.

1) 1 B3 . using the Bounded Convergence Theorem. so yields with probability 1.1).6).2) (14.3) (14. 13.á ‚   D   ë ®ê ä  † ã ß î ê æ ¼ ì 'â í â ö€ ‡ … ¢Fï ˆ–†m„ ô öӀFˆ–†…m„ ö € ï ‡ … ‰Š ô ‰Š ‘ ÓFŽˆ–†m„ ô ö¢€Fï ‡ˆ–†m„ … –†… † •† àô ‘ ‰Š lö¢€F ‡ˆÓm„ •Žˆ–†m„à ô ö Fï ‡ … ô € ⠂ ‰Š ‚ ö ‰Š ‚ öH† € à ‡ ô … ‰Š ‰Š ‚ ÓF ˆ–†m„ • †  C  à ô ‰Š € Œ  3 Œ 4   6 7 ¼ 3 ä â€ á £ ‚ 4   Œ ï yƒ s • © ä ö þ Œ  ¼ ‚ †à ÷ m£ ô ú ù ‰— •‹ ö '‚ G à ô y’ø ‰Š        ( ) ì & ' Œ‹ 'e‚ 0 $ # % ä •  Œ ©    • We have 150 if if    ‚ à à ö ¨‡yàwü –†… ÷ þ ¤ywý–†aœ÷ £‡à ü … „ ô ƒ úù y’ø ì º äƒ ‡ ü þ ‰ñò ‚ yà ¢–†… ÷ m£ û y’ø úù       ê l¼ þ ©   We use the Reflection Principle below (see Fig.: .4) (14.t.f. to get . Let in (14. we obtain . Brownian motion reaches level is infinite. We have the m.g. ‚ à ‰— ƒ A ' 6 7 ä ¼ ‚  ö †à ÷ þe£ €œ÷ ô     ¥¦ ¨§ ¡ Conclusion. The expected time to reach level (14.r.3) w. 4 @ € ä ¼ à ‰— £ 8 9 3 4 5 ä ¼ ‰—   ÷þ ‚   3 ( ) ì & ' ‹ e‚ ‰— ƒ ÷ 1 2  £  úù y’ø ‰— ! "     £ ¤ ¢ Let † Let to get $ # Letting Differentiation of (14.

τ we get t shadow path 151 Brownian motion . f Œ 4   6 7 ä † î‡ º ä  ¼ ‚ @Hˆ™†… œ ¼ ì í ‚ 4 @  U 4 @ ä ¼ ‰— Laplace transform formula: † † ‡‡ Œ ¿ˆ–†… … ² £ ‚ d V V d eV b ` E T E c a V X Y V  V W Q »­pº« î‡ ‡ • ¤' … ² ‚ ˆ™†… í † ‡ ë ®ê ä † ö • ê è ¼ ã'€ â á ‚ H† à ô ‰Š ‚ ˆ–†… ë è ê ä ã ö Œ îê ¼ ì í 'â á ‚ H† à ô ‰Š â `  U E T R S  ‚ î •º E G ‚ 6 E CHAPTER 13. Brownian Motion Using the substitution º 6 F H PI F then Density: which follows from the fact that if x Figure 13.6: Reflection Principle in Brownian Motion.

152 .

and 14.Chapter 14 The Itˆ Integral o The following chapters deal with Stochastic Differential Equations in Finance. 1993. is a continuous function of . ‰ £ g wvu t  r s p q h i (See Fig. If . has the following properties: ‡ ~–†… ‚”ƒ ‡ … ~–†a„ ‡ ƒ a„ … U .1995 2. First we will consider first variation. always in the background.1 Brownian Motion x 2.3. J.normal. even when not explicitly mentioned. Technically. Stochastic Differential Equations. g c ‡ þ ¼ ™†…a„X£¤‡ –†…a„ ¢Œ%QŒ–• ‡ ™†…a„X¤‡ þ –†a„ £ … • Œ † Q%T þ † † ŒŒŒ † ‹e‚ ö ƒj‚ ‡ •yƒ …a„ · ô ‰Š õ õƒ yj‚ ‡W • … ³ƒ ‡ ‡ • –†m„ … • • .) Brownian motion. y Y  G   f 3. of a . Futures and other Derivative Securities. 14. . References: 1. Options. function . Oksendal. Hull. B. then the increments f are independent.2 First Variation 153 ‡ … Quadratic variation is a measure of volatility. is given. U ‚ )` þ ¶‡ … „ £ þ … ³ Œ †e£ ¹ † ‚ ½œ ™†auÜþ‡ ¹ ™†a„ m ‰— … ³ • ƒ ¶‡ … „ yT‚ hœ ™†au£ ‡ ¹ ™†a„ m ‰—  P y €  P y Y         p P   p 7 1. 13. Prentice Hall. Springer-Verlag.

first variation measures the total amount of up and down motion of the path. i  † U . i.1. Then the Mean Value Theorem implies that in each subinterval there is a point such that ¶ þ ¹ † • †³    ‡ † þ … Œ œ X£ ¹ –†%‡ –†…  P U  P   i U  ˆ U f – — h y f ‡ £ þ ‚ œ –†… ¤‡ ¹ –†… £   ee e ™gf ™e ¢ ¥ñò U ‚‡ … U ‡ … d† 7f  P „ U ‚ )` We then define ¶ yƒ ³ •   ƒ i For the function pictured in Fig. The general definition of first variation is as follows: ƒ i y f ’ Definition 14.1 (First Variation) Let be a partition of The mesh of the partition is defined to be     Suppose is differentiable. 14. is given by: Œµ ‚ † Q%T ŒŒŒ ö † y%Q%• þ † • •ŒŒŒ  ƒ “ y   G ˆ ‰U y ù … ˜˜˜… c™™™7f •   þ† † ‚ ·ƒ † ÿ‚ ô ñ – —  f  ‚  ” ˆ ‰U ’ ˆ ‰U  ”  ‘f † f . .1: Example function .e.154 f(t) t2 t 1 U T t Thus.. the first variation over the interval ƒ U U   U U U U ‡ … 5† 7f „ ‚ B)` † ‡ Œ @î ˆ™†… ‚ í êº º †î ‡ ‘ †î ‡‡ £ ¢'¤¿ˆ–†… ~…  ‘ @´~–†… †î ‡ Œ '¢ˆ–†… ‚ í ºê í ºí ¶ ½‡ †–… ܇ … ³ ‚½‡ þ –†… ¤‡ ™†… ³ ½‡ ƒ … ¤‡ þ ™†…  ³ ‚ ‡ … £ ‘¶ £ £¶ £ ¶ yƒ ³ • U þ Œ ‡œ ™†… £ ‡ ¹ ™†… þ ¼ þ¼ † þ Œ ‡ e£ ¹ –†… ˆ @U ‡ ~–†… † Figure 14.

The Itˆ Integral o 155 .Œ‹ 'e‚ ö  ‡ † œ X£ þ ¹ –†… ‡ –†…      i   ˆ U   ‡œ e£ þ ¹ ™†… ‡ ™†… þ Œ † ¼ ‡ † œ e£ þ ¹ ™†… ‡ ™†…     i f – — h   y ˆ U         i   ˆ ‰U ‚‡ … ƒ m oU l U    þ Œ ‡œ †–… £¤‡ ¹ †™… þ ¼    f   U U ¶ ƒ ³ • ƒ  Definition 14. Theorem 3. then . the paths of Brownian motion are not differentiable.3 Quadratic Variation and CHAPTER 14. ƒ r … ‚ ‡ … ‡ ¥•Œ m„ õ ƒ p q m l g p q7 p • µ ‚‡ … „ ƒ “ ƒ m l Œƒ yT‚   f ”  ’ † ‡ @î ˆ–†…   ˆ U f – j f y h £  ” f £ † í  ” ’ ee e ™gf ™e ¢ £ ¥ñò ¥ñò ‚   ” f þ¼ e™egf ™e e ¢ ¥ñò Œ e™egf ™e e ¢ ‡ … ƒ m eU l  ” ’  ”  – j h   f – — h y f     ‡ £ þ¹ þ ¼ ‚ œ –†… ¤‡ –†… þ ¼ y   U U – — f y h £ e™enf ™e e ¢ ¥ñò ‚‡ … ƒ m eU l † ‡ Œ @î ˆ™†…  ˆ ‰U f – — f y h £  kf † ee e ™gf ™e ¢ í ‚ ‚‡ … U „  ‡ … 5† 7f ‚ )` – — h y þ¼ f – j h ¥ñò      f Then þ¹ ‡ þ ¼ ‚ œ ™†… £ ‡ ™†… þ ¼ y U U or more precisely. because ƒ 14.2 (Quadratic Variation) The quadratic variation of a function on an interval is U ‡ † œ ˜£ þ ¹ –†… ‡ –†…       i ˆ U   ‡ † þ • œ e£ ¹ –†… ‡ –†…    P i ˆ U In particular.44 and ô ‰Š Remark 14.1 (Quadratic Variation of Differentiable Functions) If is differentiable.

To simplify notation. then | € €  £ f Thus we have ) u … { • z w x . so  s     ‡ † þ … ¶‡ … „ £ þ … ‡ † þ … Œ C X£ ¹ –†A£ ˆC –†as܇ ¹ –†a„ ³ ‚ œ X£ ¹ –†A£   Œƒ T‚ ‡ ƒ  P  P £ f    f £ u … ¥ñò f – — h y ee e ™gf ™e ¢ ¶‡ † þ … Œ ½œ e£ ¹ –†F£  s   f s – — h Œ ‚ . the terms   and ‡ X£ þ ¹ –†… £ †    s ‡ e£ þ ¹ –†F£ † … y y are independent. set  ts f u þ¼ ‚ y ‡ ™†aX¤‡ þ ¹ ™†a„ … „ £ …   ¶ yƒ ³ • ƒ  be a partition of Proof: (Outline) Let . Define the sample quadratic variation ö •ŒŒŒ † † yQ%Q• þ † • † f‚ ô y € f ’ Then ‚ ³þ ¼ A £ ƒ f u For Consider an individual summand We want to show that 156 . ” ’  ” Œ ŽŒ ƒ ™  ‡ ƒ ƒ £ £ f f u u ⠕ƒ yj‚ ‡ … … ‰— { • z  ‚  ‚ ƒ ’     f – j h  ‚  ‚ ’  y   f – — h Œ Ž ⠂ ‡% ˜£ þ ¹ †–… † þ¼ ⠇ †e£ þ ¹ †–… þ¼ ⠂   y               } ‚ ‡ … ‰— |      P  s  i P   | ts     ¶ ‡ †X£ þ ¹ –†…ÿ‘ ‡ X£ þ ¹ –†… £ ‡ e£ þ ¹ ™†… ³ þ ‚ † † â ¼ ¶ ‡œ †e£ þ ¹ –†…T‘ C e£ þ ¹ –†… £ ‡ † â ¶‡ † ½% X£ þ ¹ –†A£ … f – — h } ~ y f – — h y  P    s f – — h   ³ { • z ³ m ‰— þ ¼ ‚ þ¼ ‚‡ ƒ y   y s ‚ ð v þ Œƒ ¶ † yj‚ ½‡ X£ ¹ –†… £      P    P  s f – j h ƒ £ f ³ þ ¼ ‰— ‚ ‡ y  P   u … ‰— This has expectation 0. so (if is normal with mean 0 and variance .

consider the equation This says that quadratic variation for Brownian motion accumulates at rate 1 at all times along almost every path. this approximation becomes exact.4 Quadratic Variation as Absolute Volatility   ƒ i ƒ On any time interval . Brownian motion has absolute volatility 1. The Itˆ Integral o £  ƒ 157 Remark 14. Furthermore.CHAPTER 14. we can sample the Brownian motion at times   ƒ “ and compute the squared sample absolute volatility h y This is approximately equal to l ƒ ƒ As we increase the number of sample points. and we have the approximate equation    P þ … ‡‡ … „ £ þ … Œ ƒ ¶‡ † yj‚ ½œ e£ ¹ –†F£ Hœ –†mX¤‡ ¹ –†a7„… a ‰— ³  P   Œƒ T‚ ‡  P   ƒ £     f   u … ¥ñò f   £ ee e ™gf ™e ¢  P    P   £ f … { • z £  ” ’  ” As . so .2 (Differential Representation) We know that We showed above that { • z which we can write informally as 14. † þ ‡‡ … „ £ þ … • e£ ¹ † Hœ –†mu܇ ¹ –†m7„… ‡ ˜£ þ ¹ ™†… † ‡ † þ ¶‡ † þ … ‡‡ … „ þ … Œ % X£ ¹ –†… ⠂ ˆC e£ ¹ –†F£ ~' –†aX£ ‡ ¹ ™†a7„… ³ is very small. þ £ Œ'‹e‚ þ £ Œ yƒ “ ‡‡ … „ £ þ … Œ Hœ ™†aX¤‡ ¹ ™†a7„… þ ¼ ƒ ƒ       ƒ @…     þ ‚ † Q%T † ŒŒŒ ƒ ƒ † ‡ … î ‡ … Œ 'î ‚ ~–†a„ ´~–†a„ î y €   þ £ ¶ þ ‚ ˆ‡ … „ £ ‡ … „ ³ ‹  G † • 'î ‹ ƒ „  ƒ   m    † f í l ‚ F ‚‡ … „ f  – j ƒ “   f ƒ þ £ † ‚ þ ƒ m    ‹ ƒ   ƒ m ƒ l  ¶ •þ ³   ‡ e£ þ ¹ –†… †   When is small. ƒ ‡ u ƒ . In other words.

i. because the paths of Brownian motion are not differentiable.r. is -measurable. (see Fig. the increments .. is square-integrable: ‡ † … ‡ ~–†… ‡ … ˆ™†3 ‡ 1. î‡ ‡ … ‡ î‡ … Œ ¤† … † y º ‚  … ¤† y í ºí .2). “ ‡ ƒ c† • ˆ™†…  ‡ … ƒ “ † • ~–†a„ . where (i. ‡ þ ¼ –†mu¤‡ –†m„ y%Q%• ‡ þ –†mu¤‡ –†a„ • ˆ™†au£ ‡ þ –†m„ … „ £ … •ŒŒŒ … „ £ … ‡ … „ …  † … “ ‡ … ƒ  ´† • ˆ™†3 ‡~–†… † Œ%%Œj þ †  † Œ ‡ ˆ–†… ~–†a„ ‡ …  2. ‡ ™†3 …  ‡ ‡ … ˆ†–m„ ‡ … ˆ–†m„ The functions and can be interpreted as follows: ‡    P Assume that is constant on each subinterval elementary process. . every set in is also in . then y € y €   y € 3. is ‡ The integrand is .6 Itˆ integral of an elementary integrand o ƒ i y f ’ Let be a partition of f † Think of as the price per unit share of an asset at time . and the ‚† an  . We call such a ˆ G ‡ ~–†… Remark 14..158 14. with associated filtration We want to define the Itˆ Integral: o This won’t work when the integrator is Brownian motion. 14.5 Construction of the Itˆ Integral o The integrator is Brownian motion following properties: . a differentiable function) If we can define ˆ ‰U ˆ c ‡ f ˆ c U ˆ c ‡ Œ µ Œ “ yƒ –† ƒ @… ¶ þ ¹ † • †³ þ Œµ ‚ † ŒQ%ŒT † † ·ƒ Œ ‚ ¶ yƒ ³ • ƒ “ U ˆ c • y Y ‡ … „î‡ … •  m‚¤† y  S   G ˆ c †î ‡ @´~–†…  ˆ c ‡   ‡  † f í ‰— ºí f ‡ ‚ ~–†… ‰ f ö † y%Q%• þ † • † ÿ‚ •ŒŒŒ ô ‡ … ~–†y ‡  ‡ 2.e. For are independent of  T   .t. 14.3 (Integral w.e. ‡ ˆ™†… ‡ Q7’… † e  ’ 1. is adapted) -measurable is a differentiable function.

  Then the Itˆ integral o 14.7 Properties of the Itˆ integral of an elementary process o f f † †–y … –†y … ‡ – f –†y … ‡ ¥‰‰  P   ‡ f In general. ‡† y … … ‡ ˆ–†… º í ‚ f ‹  Ž ‡ ~–†… ‰ can be interpreted as the gain from trading at time . t1 δ( t ) = δ( t 1 ) -measurable. The Itˆ Integral o 0=t0 δ( t ) = δ( t ) 0 Figure 14.2: An elementary function . ‡ ‚ ~–†… ‰ δ( t ) = δ( t 2 ) t2 δ( t )= δ( t 3 ) t3 t4 = T and held until 159   . this gain is given by: as the trading dates for the asset. ‡ Think of y € Think of trading date as the number of shares of the asset acquired at trading date .Œ ‡ … î ‡‡ † a„ ¿ … ˆ c ˆ c Œ f  e c c ˆ ‡ f ‡ … î‡ † a„ † … ˆ c ˆ c ºí ‡ ‚ ˆ™†… ‹ ‡ … „î‡ … •  m‚¤† y ˆ c ˆ c ‡ ºí  ‡ ‚ ˆ–†… ‰ ‡ ~–†… ‰ • † Linearity If Adaptedness For each is ‡ ~–†… X         Œ † † þ† † † † þ† † ƒ   t ¶ … £‡ ‡ … ‘¶ „ þ … … ‡ Œ ½‡œ –†m„u¤~–†…a„ ³ % †™3 y½‡ –†…aX£ ‡ ¹ –†a„ ³ ‡ –†y þ ¼ ‚ ~–†… ‰ þ¹ † † þ … ‘ … „ £ þ … … ‡ ¶ þ … „ • ¶½‡ ™†a„u£ ˆ–†…m„ ³ ‡ ™†…3 ‘‚½‡ †–mX£¤‡ ™†…a„ ‡ –†y 3¶½‡ –†as¤‡ –†a„ ‡ ¶ þ „ £ … ³ þ ‘ … „ £ þ … ³ • ½‡ –†…mX¤‡ˆ™†a„ ³ ‡ –†…y 3¶½‡ –†as¤‡ –†a„ ³ ‡ ­ ´¸ « ¦ £‡ … • ¶ ‡© –†D…§ m„ ¥ ¤~–†a„ ³ ‡ † ‡ ~–†… ‰ þ¹ † † ‡ … % ™†3 † †%Q%• þ † • † •ŒŒŒ     ‡ Š y y y ‡ f – y h          ‡ Š   ‡ Š ‡ Š f f f – f f CHAPTER 14. if   ‡ ‰ ‰ ‰‰ ¨ ‰‰ § then .

‘ ‘ ‘ ‡ ‡  ‘ ‘ ›› ›› ˆ‡ ‡ ‘ ‘ ‘ ‡ ¶ … „ £‡ … ½‡  –†mX¤%7’m„ ³ ‡ ‡ „ ¶‡ H‡ –†…as£ h%7’… ‡ þ ¹ –†m„ m ‰— ‡ … ³ … ‡ … „ £ þ … … … Œ H‡ ™†aX¤‡ ¹ –†m7„%‡ –†y y y y ‡ f – y ›› ™†3 ‚ … ™†3 ‚ ½Q7’… … ¾‡ ¯‡ þ ¼ ‚ ± %7’… h ‘ ›› H‡ ‡ y y y ‡ f – y –†m„X¤‡ þ ¹ –†m7„%‡ –†y ˜ ‰— … £ … … … –†…as܇ þ ¹ –†…a7„Q‡ –†y „ £ … … þ ¼ ¡ž ‰— h ‘  P   ‡  y y ‘ c ‘ c ‘ ’ ‡ “ y þ¹ ¶ „ ‡ … ‡ … ½‡œ –†…aX£¤~–†a„ ³ œ –†y ‘ ½‡ ™†aX¤‡ þ ¹ –†m„ ³ ‡ –†y þ ‘ ¶ … „ £ … … ¼ ¶‡ … ½D †–m„X£¤‡ þ ¹ –†…m„ ³ ‡ –†y ‘ ˆ‡ –†aX¤‡ þ ¹ –†m„ ³ ‡ –†y þ ‚ ~–†… ‰ … ¶ … „ £ … … ‡ ¼ ‘ – h y ‡ y  y y ‡ f – y h ‘   ‘ ’ ‘ c þ¹ †    †  160 Figure 14.e.. 14.. ¶þ ¹ † • †³ † † p ’ ¶ þ ¹ † • †³ p ’  † ‘ c †   †    P ‡  y ¶‡ … „ £ ‡ … ‡ … ¶ … „ £ þ … … ‡ • ½œ ™†aX¤~–†a„ ³ C –†y ‘ ½‡ –†mX¤‡ ¹ –†m„ ³ ‡ –†y þ ¼ ‚ ˆ–†… ‰ y y ‡ f – y  h We prove the martingale property for the elementary process case.. t k t t k+1 are in different (See .3: Showing and in different partitions. We treat the more difficult case that and and subintervals. We compute conditional expectations: †  ’  r ƒ ‡ ˆ–†… ‰ Write is a martingale. t l s t l+1 . † ’ and î … ‡ Œ ‡ …m„‚H‡† y º ‚ ˆ–†… ‰ í ˆ c ˆ c ‡ f   Theorem 7...3). Martingale is a martingale.45 (Martingale Property) Proof: Let be given. i. there are partition points and such that Fig.

We show that the third and fourth terms are zero. –  f ‡  y s – y h –—  h ‡ –†… y y ‚ ‡ ‚ ~–†… ‰   s  ‡ –†y …   ›› ›› Hœ ™†aX¤~–†a7„QC –†y ˜ ‰— ‡‡ … „ £ ‡ … …‡ …    P ‡ ‘ – h y y y ‡ þ¹ ›› ›› ~‡ –†…m„u£¤‡ þ ¹ –†…m7„%‡ †–y þ ¡ž ‰— ‡ … … ¼ y  ‡ Q7’… ‰ Each y s Proof: To simplify notation. 161 . and different . so are independent.46 (Itˆ Isometry) o These first two terms add up to CHAPTER 14.ˆ G ˆ c ‡ f   • ˆ G ˆ c ‡ f – y   › €•  h y y y ‡ f – y    h î‡ ¤† … º ‰— ‚ º í î‡ ¤† … ‰— ‚ íº  ‡ e£ þ ¹ –†%‡ –†… ‰— † … ‚ ›› ‡ … „ £ … ¾ C¾ ‡ –†… ›› ~‡ –†as¤‡ þ ¹ –†a7„… ˜ ‰— ‡ –†… ™ ‰— ‚ ˜ y   y y y ‡ f – y    h ¶ ‡ –†… m ‰— ³   y s  y ‡ f   & ‡ y šh s y š s  ‡  s  y   ‡   f – y h  ‡ ‰ ‚ ~–†… " ‰—   y Œ ‡ –†y ‡ –†y … … y š â   ‘ ˜™ jy • B” y y y ˆ T ˆ ’ ‡ f § ¶ ‡© –†as¤¦ ‡ þ ¹ –†a„ ¥ ³ ‡ –†y … „ £¨ … … ‡ ‚ ˆ™†… ‰ † ‚† î‡ ‡ ‰ Œ ¤ … º ‰— ‚ ˆ™†… " ‰— í ‡ f – y  h     f –       P ‡ f –  ‘ – h y y y y ‡ y  ‘ – h y y y y ‡ y  ›› I°±¯ %h… ›› Hœ ™†au½Q D–†§ ¦ … ˆ–†m„ % –†y ž ¡ ‚ ¾ %7’… ‡ ’ ‡© ‡ … „ £ ¶ ‡ ‡ … m ‰— … ¥ ‡ … ‰— ‡ ³  þ¹ ›› ±¯° %h… ›› H‡ –†…mu½‡ –†… ¨¦ ‡ þ ¹ –†a„ a ‰— ܇ ™†3 ž ¡ ‰— þ ‡ ’ ‡© „ £ ¶ § … ³ …¥ … ¼ ‚ þ ¹ ›› ›› ‡ … ‡ ¾ ‡Q7’… ›› ¾ ‡ †–… ›› H‡ –†…a„s£Ü‡ þ ¹ –†…a7„Q‡ –†…y ˜ — –‰™ ‰— þ ±¯ Q7’… ˜ ¼ ‚  Since the cross terms have expectation zero. The Itˆ Integral o has expectation 0. . assume Theorem 7.

Proof: Fig. over ¶ yƒ ³ • . Let be a process (not necessarily an elementary process) such that ƒ  p … is -measurable.8 Itˆ integral of a general integrand o ‡ ƒ Fix .4 shows the main idea.162 path of δ 4 path of δ 0=t0 t1 t2 t3 t4 = T ƒ i 14. Œƒ † ‡ … £‡ yT‚ 'î ˆ™†3 ¤ˆ™†… þ ìy ö ô ‡ … „î‡ … ˆ–†m‚¤~–†Ó y —‡  † f ‡ Ÿ í y —‡ ì ¥ñò £ y † ¢ f y —‡ í  ‡ … î‡ … ‚ ~–†a„ ¤ˆ™†3 … ‚ ‡ ý‰ ¶ yƒ ³ † • f ì ¥ñò í ‰— ƒ y ‡ £ y ¢ † f í Œ  S  †@î´~–†…  ‡ ‰—  ‡ ~–†… ˆ™†3  ‡ …   ‡ f œ †    ‡ ƒ ï . We now define ‡ … î ‡ Œ ~–†a„ ´~–†… – žy y —‡ Theorem 8.47 There is a sequence of elementary processes   † such that | ‡ Figure 14. . 14. In the last section we have defined for every .4: Approximating a general process by an elementary process .

1 () Consider the Itˆ integral o We approximate the integrand as shown in Fig. Example 14. is ‡ … î‡ … ‡ Œ  a„ ¤ y º ‚ ˆ–†… ‰ í -measurable. 14. o † ‡ … î ‡‡ † a„ ¿ … ˆ c î … ‡ Œ ‡ …m„‚H‡† y º ‚ ˆ–†… ‰ í ‡ ˆ–†… º í ‚ then ‹  Ž ‡ … î‡ † a„ † … ˆ c ˆ c Œ ºí f ˆ c ˆ c ‡ ‚ ˆ™†… Œ  e c c ˆ ‡ ˆ c ‡ ‡ … †y… ‹  f ‡ … „î‡ … •  m‚¤† y f ˆ c  ˆ c ‡ î‡ ¤ … º ‡ ~–†… Adaptedness.5 Ò Ð ·Í Ð ·Í Ï Ï í « ¬ ­ Ž « ¬ ¨ © ª ˆ T ˆ ’ ‡ f Itˆ Isometry. If and Martingale.CHAPTER 14. Continuity. . is a martingale.    ‡ §   £  ‡  † f £ ‡ Ÿ ‡      ‡  ‡ § ‡  ˆ c ‡ f y —‡ y —‡ y —‡   † f † f (Itˆ Isometry:) o   ¥ P   £ ‡  £ y —‡ † f † f ¤ " ‡ ~–†… ‰ ºí f     ‡ ‚ ˆ–†… ‰ r ‡ ~–†… ‰ †   œ Ÿ ƒ ¦ ‰— ‚ ‡~–†… ‰" ‰— ‡ ~–†… ‰ ‡ ~–†… ‰ £ —   d ƒ    y   { ¢• z ¦ Š d ¡ ‡   . Linearity. For each . – žy ˆ ’ ƒ y which is small. is a continuous function of the upper limit of integration . Then 14. The Itˆ Integral o 163 The only difficulty with this approach is that we need to make sure the above limit exists. This guarantees that the sequence þ ìö … ¨‡ ý ‰ ô ‡ … £ ‘ @î ˆ–†y ¤~–†… † ‡ … £‡ ‡ ‘ ‡ ’ ‘ H… … • †@î ~†–y ¤‡ˆ–†… â â í ‰— â í ‰— ⠆ 'î ¶ ‡ˆ–†… £Üˆ–†…y ‘ ˆ–†y £ ˆ–†… ³ ‡ ‡ … ‡ ‚ í ‰— †@î ½~†–… £ ˆ–†… ¶‡ ‡ ³ í ‰— ‚ ‡ … ¶‡ ~†–a„ î ½~†–… £ ‡ˆ–†… ‡ £ … ³ í ‰— ‚ H‡ … ‰ ܇ ý ‰ … has a limit. square-integrable process. Suppose and are large positive integers.9 Properties of the (general) Itˆ integral o Here is any adapted.

we denote By definition.5: Approximating the integrand ‡ … † a„ ˆ c T/4 2T/4 if if Ì ¢ ¶ µ ´ ·uv³±D ² « ° à if Ò ¤ ´ ² « ÂÁÀ° » ¿ ¨ » dº ¾ ½ ¼ ’‰¯ Í ÒÒ ¨DÒ Ï Ð ·Í à ýÐ »Ã·Í Å Ï ¶ µ ´ ·u” Å ýÐ Ï ¯ ® so í We compute To simplify notation. 3T/4 with .Î ½ Ô 7È ª ÓÈ É Î h ½ ¼ @¢¯ ª ÓÈ É ÒÐ Í È ½ ª ÓÈ É h ½ ½ ¼ ‰¯ Õ È Õ Ô ÓÈ –ÍÏ Í È Î Õ È Õ¯ Í Å Õ ½ Õ ½ ª 7È É h Í Í È È Í h Ø Í Å Õ¯ ½ ¼ @¯ ª ÓÈ É Î h ½ ¼ ‰¯ Õ Ö ½ ¼ @¢¯ ª jÖ É Õ È Í Ø ½ ª ÓÈ É h ½ ¼ @¢¯ Ô 7È Í Í Ø ½ Ô 7È Í h h ½ ½‰¯ ¼ Î Õ Ø Í Å Õ¯ Õ ½ ª 7È É h ½ ¼ @¢¯ ½ Õ Î ½ Ô ÓÈ ª 7È É h ½ ¼ @¯ ½ Õ Í Õ ½ ª ÓÈ É Í Í È ÔÓÈ ½ Õ Í Å Ð Í Õ È ­ Ž « ¤ »ÍÏ ½ ¼ @¢¯ Î ½ Ô 7È ª ÓÈ É h ½ ¼ @¢¯ ¶ ´ ÒÐ Í È –ÍÏ Í È Ñ ÇÌ Æ Å¦ ž¯ Ã Ï pÆpÇ ýÐ ·Í Ð ×Í Å Ï « ¤ ª ¨ Ò Ë Ê Žx Í Å Í Ð )È Ò ¾ Ì Ï¶ ´ Ë Ê nr Í Î ÇÌ ´ ¶ Ë Ê r Ì v¶ ´ Ë Ê gr Ð ¨ Ø Ï Í Ÿ ¨ ͘ Í ¨ ª 7È É h ½ ¼ @¯ Æ Çž¯ ÅÄ Ã « ¤ ­ Ž « ¤ ª Ï pÆpÇ ýÐ ·Í Ð ·Í í Å Ï ¨  ‰ ‰ ‰‰ ¨ ‰‰ ¥‰‰ § « ¤ ¯ ¾ ½ ¼ €@¢¯ Ì ¢ ¶ µ ´ ¹ ·u7§³§¸·u´ ² « ° ¶ µ | ‡ ¶ yƒ ³ • ƒ  164 Figure 14. . over T .

4 (Reason for the ƒ   þ í ¶ Ê x Ì §¶ ´ ͘ Therefore. because ƒ   In contrast. It has to be term) If is differentiable with . then 165 (Itˆ integral is a martingale) o .48 (Quadratic variation of Itˆ integral) Let o ƒ ‡ … „î ‡ … T‚  m‚¢ m„ ˆ c ˆ c † f f í ‰— þ í     f † Œ ›› ‡ … þ ‚ ››  … þ ‚ ‡ ˆ c U U f ˆ c í ‡ î‡ ‚ † … ¤† … ˆ c U † U U f ´ ‚ Ï Ð Ï Í yÐ ·Í Ð ·Í í Å Ï Õ Õ « ¬ ½ ­ Ž « ¬ ª ¨ Ø × ÇÅ Ð Ï Í Å ¾ ´ ” Õ Õ ½ ª 7È É h ½ ¼ @¯ ½ Õ Î Õ¯ Õ Ì §¶ ´ Ë Ê Žx Í Î vÌ ´ Ð ¨ Ø Ï Ë Í  ª 7È É h È Î ½ Ô ÓÈ Í ýÐ Í Å È ½ »ÍÏ Í Remark 14. ½ ¼ @¯ The extra term there. for Brownian motion.10 Quadratic variation of an Itˆ integral o and use the definition of quadratic variation to get comes from the nonzero quadratic variation of Brownian motion. The Itˆ Integral o but Œ Ž þ ‚ ‡ … „ þ ‰— ƒ   ƒ     Then ‡ Œ † … º ‚ ~–†… ‰ í   m l 14. we have Ë Ê gr Í ª 7È É h ½ ¼ ‰¯ Let ¶ or equivalently CHAPTER 14.ˆ G ˆ c ‡ f ‡ … î‡ … ‡ Œ  a„ ¤ y º ‚ ˆ–†… ‰ í ˆ ’ ˆ c ‡ f ‡ … „ î‡ … Œ þ £ µ ÿ¤‡ … „ þ ‚  m‚H† a„ ƒ   ƒ     ˆ c ˆ c † ƒ     ˆ T î‡ ‡ ¤ …  … ˆ c ˆ ’ U † ˆ ‰U ƒ T‚ ‡ ƒ … U Ò ¤ Ì)´ Ë sÊ Ì ´ ¶ ª ÓÈ É h ´ Õ ½ Î Ò ¾ Õ Ë Ê r Ð D Ø Ï Í Ÿ Ñ Õ Í˜ Î Õ ½ Î ½ ¼ @¢¯ Ð Í È ½ Ô ÓÈ »ÍÏ Theorem 10.

Let for . To simplify notation. ) in the Brownian motion.e. the instantaneous absolute volatility of is . we can write the quadratic variation formula in differential form as follows:   Proof: (For an elementary process ). . We have ‡ î It follows that   P † ‡ … Q –†y   ‡ ˆ so Then Let us compute .ˆ T ˆ ’ ‡ f £ ”  ”  ” £ ’   Ú ˆ T ˆ c ‡ f – j h   › Ú ¤ y    P   ‡ f – j h £££££ †yyy®£ î Œ ¤‡ … º í ƒ º  … ‚ íº þ ¼  ‡œ †e£ þ ¹ †–%‡œ †™… þ ‚ ˆ–†… ‰ … ‡ ¼   y m l  P  P   ‡ £ T  ”  ” £ Ù   – f y   y y  P ‡ h   £ f – y ££££ yy†£ ‡ † þ …‡ Œ œ X£ ¹ –†%Q –†… ƒ ¶ ˆ‡ 7’…a„X¤‡ þ ¹ ha„ ³ þ % –†… ‚ £ ’… ‡ ¼ ¶ ’ ½‡ h… ‰ £‡ þ ¹ 7’… I‰³ þ ‚ ’ –†… ‰ ¤‡ þ ¹ ™†… ‰ ‡ £ ¼   y y     h £ m l m l y y  P ‡ • ˆ c   ‡ y 7’… ‰ £ ‡ þ ¹ 7’… ‰ y  f   › €•   £  ¢  £ ’ ‚ † ‡Q –†… ‰ £ ‡ þ ¹ –†… ‰     m l m l m l  P  P f – — h y m l m ‡ ‚ ˆ–†… ‰ ‡ l y € þ¹ †  P  ‡ … ‚ ~–†y ‡ ¶ … „ £ þ … • ½‡ ’7mu¤‡ ¹ 7’m ‡ … ³„ % –†y ‚ ¬ ‡ … „î‡ …  m‚¤Q –†y ‚‡ í¬  þ þ Œ ¹ † ‚ ’ %%j ’ ŒŒŒ ö ’ y%Q%• þ ’ • ’ ÿ‚ •ŒŒŒ ô ‡ £ þ Œ ¶½’ –†… ‰ ܇ ¹ –†… ‰ ³ þ ¼ † ‚† öy † •Œ%Q%• þ † • † f‚ ŒŒ ô † ‡ … î ‡ … Œ 'î ‚ ~–†a„ ´~–†a„ f Ù y € f ’ † †î ‡ Œ '¤~–†…   ‡ ‡ … ˆ–†y ‡ †… ˆ c ‡ This holds even if is not an elementary process. Let be a partition . i.e. This is the absolute volatility of the Brownian motion scaled by the size of the position (i. The quadratic variation formula says that at each time .. Informally. assume ‡ ‡ î‡ ‚ ~–†… ‰ ¤~–†… ‰ î ‰ ‡ Compare this with 166 be the partition for .

Integral Forms) The mathematically meaningful form of Itˆ ’s foro mula is Itˆ ’s formula in integral form: o 167 î ‡‡ … Œ ¤H m7„… î ‡‡ … Œ ¤H m7„… ˆ T ˆ T ‡€ h… Œ U ˆ c ˆ c H ˆˆ ™‰U ˆˆ ™‰U ­»pº«´¸ »­º ¢¸ « © † ¨¦ ‡ ‡ … „ þ ‘ ‡ … „ î ‡ ‡ … „ ¨'îD¥ H~–†ah… f܈–†m‚¤H~–†ah… § º í º í f f ‡‡ … ¿ˆ–†m7„… H     ‡ „‡‡ … • ˆ–†… HHˆ™†a7„… ‡ … „  … ˆ–†m‚¿Hˆ–†m7„… †î ‡ „‡‡ … '¢ˆ–†… ¿Hˆ–†m7„… ­ « § þf‘ý‡† a„ î¤H‡† …a„7… … ‡ ‡© … D £ ‡ ‡ … H‡ m¦ 7„… ¥ ¤H~–†a7„… º í ‚ ƒ þf‘ý‡† a„ î¤H‡† …a„7… … ‡ ‡ … H‡ m7„… ¤H~–†a7„… £ ‡‡ … º í ‚ ƒ ˆ U ˆ ˆ c ˆ c ˆ ‰U ˆ   ˆ U ˆ c ˆ c ˆ U ˆ ‰U † ‡ ‡ … @î ‚ ¿ˆ–†m7„… î ˆ ‰U ˆ ‰U ‡‡ … „ ‚ H~–†ah… î ‚ f f U ˆ ‰U U ‡‡ … ‚ Hˆ™†a7„… î f Û U U U U U ‡ … ~–†a„ ‡ … ~–†a„ . then the ordinary chain rule would give function.1 (Differential vs. Integrating this. so the correct formula has an extra term. This is Itˆ ’s formula in differential form. If which could be written in differential notation as However. namely.1 Itˆ ’s formula for one Brownian motion o We want a rule to “differentiate” expressions of the form .Chapter 15 Itˆ ’s Formula o 15. is not differentiable. we obtain Itˆ ’s formula in integral form: o o Remark 15. where is a differentiable were also differentiable. and in particular has nonzero quadratic variation.

and the error is of the order of . we write: £ þ ¹ 7€…  @ In the general case. the type used in freshman calculus. For paper and pencil computations. Taylor’s formula to second order is exact because X is a quadratic function. ‡ Œ % 7€…  ‰ ˆˆ ™‰U   ‡ € Q s£ þ ¹ h… fý% 7€… œ X£ þ ¹ h… ‚ œ 7€… ¤‡ þ ¹ 7€… € þ ‘‡ ‡ € € ‡ £  @ U  @ ¶ yƒ ³ • ‡ þ ¹ –†a„ ³ þ þ @ˆC –†as¤‡ þ ¹ –†m„ ³ œ –†m„ þ ‚ … ‘ ¶‡ … „ £ … ‡ … ¼ ¼ –†m„ ³ þ þ ‘¢H‡œ †–m7„… ½œ –†mX¤‡ þ ¹ –†m„ ³ þ ‚ … ‡ … ¶‡ … „ £ … ¼ ¼ ¶½¿‡C –†…ah… ÜH‡ þ ¹ –†m7„… ³ þ ‚ ‡ „ £‡ … ¼ ‡ ƒ … „ þ ‡ … „ þ ‚ £ ‡ … H‡ ƒ m7„… ¤H‡ m7„… £‡ … öy † †%%Q• þ † • † ÿ‚ •ŒŒŒ ô ƒ ï    ‰ ˆ ‰U  ‰  @  @ U  @ U  ‰  @ Let be numbers. 15. Using Taylor’s formula.168 This is because we have solid definitions for both integrals appearing on the right-hand side. The second. the more convenient form of Itˆ ’s rule is Itˆ ’s formula in differo o ential form: ˆ ™ˆ U ˆ U U There is an intuitive meaning but no solid definition for the terms and appearing in this formula. the above equation is only approximate. This formula becomes mathematically respectable only after we integrate it. defined in the previous chapter. o is a Riemann integral. The total error will have limit zero in the last step of the following argument. so that † @î ‡ … ‡ … ~†–a„ î • ¿‡ˆ–†m7„… î † î ‡‡ … ‘ ‡ … î ‡‡ … Œ @¤H~–†a7„… þ ý~–†a„ ¤H~–†a7„… U ‡ … „ î ‡‡ …  m‚¢H a7„…   ˆ c ˆ T ˆ ’ ˆ ’ ˆ @U ˆ f î ‡‡ … • ¤H a7„… º í ˆ U º í f ‡‡ … ‚ Hˆ–†m7„… î   € þ ‚ 7€… ‡ þ¹ € € • U ‡ €  . is an Itˆ integral. ‡‡ … „ ¿C –†ah…  P ˆˆ ™‰U   ¶‡ … „ £ ½œ –†mX¤‡ þ ¹ ¶‡ … „ Œ ˆC –†aX£      P   ƒ   P f – j h y f   – — h y      P  P ˆ ‰U y Y   U       f   U     ƒ U ’   f f f ƒ – — – — – —   h h h y y y U ƒ Fix and let be a partition of .2 Derivation of Itˆ ’s formula o In this case. Taylor’s formula implies Œ‹ ‡ '‚ 7€… ˆˆ ™‰U ‡ • € ‚ †7€… ˆ ‰U   Consider . The first.

1 (Geometric Brownian Motion) Geometric Brownian motion is û 3’ø úù ‡ ƒ… Ü ‡ ‚ ˆ–†… Ü U f † f † í í ‚ ‡ … £‡ … ‚ H‡ ƒ m7„… ¤H‡ m7„… ƒ U U  ” ’  ” where Þ Define and Geometric Brownian motion in integral form is Ü ‘ @H~–†… † î‡ ‡ ‚ ˆ–†… î Thus. 169 . Itˆ ’s Formula o Definition 15. Geometric Brownian motion in differential form is   €   þ e£ Þ º ‚º U ‡‡ … Œ H~–†a„ • –†… U ‡ ‚ ~–†… Ü Ü ‘ € ”y € ‡ ƒ… ‡ ‚ †€ • ™†… U û y’ø úù According to Itˆ ’s formula.3 Geometric Brownian motion This is Itˆ ’s formula in integral form for the special case o CHAPTER 15. o Then so and to obtain are constant.‡ … î‡ Œ  a„ † … ˆ ’ ˆ c Ü € f º í ‘ î‡ ” ¤ … ˆ T ˆ ’ Ü Þ f º í € Š ‘ ¢‡ ƒ … Ü Ü ‡ ‚ ~–†… Ü Þ Ü ‡ … î‡ • ~–†a„ ¤ˆ–†… † 'î   U €   € H U ‡ ~–†…a„ –†… ‘ @H~–†… ‚ ‡ † î‡ þ ‘A„ î ‘ @î ‡ þ £ … ‚ † º © ‚„‚î ¥ § „  ‘† ’ þ ‘G„‚î  ¢'î º ‚ ‡‡ … Hˆ–†m„ • –†… î ‚ ˆ–†… î ‡ Ü € Š Ü Þ   €   Þ ¬ U   U U U Ü U € U Œ   U €  ‚ ’ • U ‚  • ¿ U U ¿   • † ¡ €   þ e£ Þ º Ý ï ƒ e € ¿   • † ¡ €   þ e£ Þ º Ý ‘‡ … „ ~–†ay € ‡ Œ € þ ‚ 7€…     ˆ G ˆ c f   ˆ ’ ˆ c ˆˆ ™‰U † † þ § þ ‘ ‡ … î ‡‡ … ‡© ‡ ¨ Œ î ¿ …m¦ 7„… ¥ fý a„ ¤H m7„… í § þ ‘‡ … „î ‡ … ‡ © … ¨¦ „ ¥ fy m‚¢ m„ ˆ ‰U ƒ   ˆ c ˆ c m l ƒ £ We let 15.

holds shares of stock.170 15. þ þ ‡ … ~–†ð ƒ ƒ 9v  1 à   ƒ ƒ . In differential notation. The Itˆ integral o ˆ G î‡ ¤† … ˆ c Ü Þ ‡ ~–†… º í ‚ f ` the Riemann integral ˆ ` ‡ … î‡ • † a„ ¤ … ˆ c ˆ ’ Ü € º í f ‘ î‡ · ¤† … ˆ G ˆ c Ü Þ º í f ‘ †‡ ƒ … Ü Ü ‡ ‚ ˆ–†… Ü ‡ ‚ ˆ™†… . the instantaneous relative volatility of is .4 Quadratic variation of geometric Brownian motion In the integral form of Geometric Brownian motion. This term has zero quadratic variation. This is usually called simply the volatility of . Ü   € Ü 15. The squared absolute sample †î‡ @¤ˆ™†…   Ü   ¶ •þ ³ € ‡‡ … „ î‡ ‚ Hˆ–†m‚¿ˆ–†…   Thus the quadratic variation of we write Ü Ü is given by the quadratic variation of Ü € Š ß î‡ Œ ¤ … ‡ … î‡  a„ † … ˆ T ˆ ’ ˆ c   ˆ c Ü   Ü € ‡ ˆ™†… º í ‚ € f ‡ ~–†… º í ‚ ‘ @Hˆ™†… † î‡ f ö •ŒŒŒ y † Q%Q• † ô ‚ Ü m Þ ¬ ß … ‚ ˆ–†… ¤~–†… î ‡ î‡ ß l ‡ ˆ–†… Ü Þ is differentiable with . In other words. It has quadratic variation 15. Stock is modelled by a geometric Brownian motion: † ˆ T f î‡ ¤ …  ˆ ’   Ü   € ‡þ …  þ £ ê í ‹   ‡ … „î‡ Œ ˆ–†m‚¿ˆ–†… † ƒ  † ƒ ƒ ƒ   Ü     ƒ € Ü € ƒ ƒ   ‘ @Hˆ™†… † î‡ ¶‡ ½% –†…  P Ü Ü þ £ £¤‡ þ ¹ –†… ³ þ ‹ ¼ ¶ •þ ³ Þ y Y ‡ ‚ ˆ–†… î   Ü Ü f f – j h y ƒ ’   ƒ   ƒ 9 Ü Fix volatility of . An investor begins with nonrandom initial wealth and at each time . is not differentiable. the above approximation becomes exact.5 Volatility of Geometric Brownian motion   ƒ  ƒ on is As .6 First derivation of the Black-Scholes formula Wealth of an investor. Let be a partition of .

and in particular. We saw above that Þ ¬ à Š  2á  ¢ and invests so that the wealth ⠃ „î ‘ý'î ’ †  þ ‘ † @î ‘† ý@î  þ ’ f‘3¶Ž„ î î î ’ f‘  þ   Ü   € â Ü € Š â   ƒ Value of an option. The differential of this value is â â Ü â G satisfying the terminal condition â f  â If an investor starts with and uses the hedge for all .CHAPTER 15. Then á ‡ … ~–†ð „î à at each  . à ƒ Ü ƒ  ‡ ¿‡ … … ² ‚ ‡ F … † ‡ H‡~–†… • –†…  ‚ ‡~–†…ð ‡ • H‡ ƒ … yƒ … ‚ ‡ ‡ • … Œ †7€… ² ‚ € µ ¢ Ü ‡ … € • –†¢ ‚ € • –†… ’ € þ ¢†€ • ™†…  C ¢†€ • ™†… º ‡  ‘‡ € ‘‡ â 'á â     €   ƒ â ⠉ Š á Ü â In conclusion. Making these substitutions. we obtain: ‡‡ ‡ … Œ H~–†… • –†…  ‚ ˆ–†ð Ü â à à To ensure that coefficients. Let denote . we obtain the â  for all . ‡ … Œ œ £ ð ‘ Ü á § Þ ¬ à   2á  ‚ ’ â   Ü   €   þ ‘  â Ü Þ Ÿ ‘º ⠍ ‚ † 'î Equating the coefficients. . Consider an European option which pays the value of this option at time if the stock price is option at each time is Ü at time . In other words. we equate coefficients in their differentials. and we are seeking to cause to agree with . we should let â be the solution to the Black-Scholes partial differential equation Œ â 'á  ‚ ’ â   Ü   €   þ ‘  â Ü á Š ‘º ‡ ~–†… ⠇‡ Hˆ–†… • –†… ‚ ˆ–†F ‡ … Ü â â â (where and ) which simplifies to ‡ ‘  • œ £ …  h ‚ ’ Ü â á § Þ ¬ â  â 'á  â   Ü   €   þ f‘  â Ü Þ Š Ü ‚ ‘º Ü â ⠇‡ ¿ˆ–†… • ™†… ‚ Ü à But we have set we obtain . then he will have ‡ … ˆ–†F  î Œ „¡ à Ü € Š † ‡‡ Hˆ™†… • –†¢ ‚ ˆ–†F … ‡ … ‘ †@ œ £ …w ‘™ ³ ‚ ˆ–†F î ‡ ‡ … Ü f  A hedging portfolio starts with some initial wealth time tracks . the value of the ‡ … € • –†¢ © ‡ „ ‡ ‡ ‡ ‡ … ‘ † î‡ … Œ ˆ–†…m‚î† ˆ†–… ˆ™†…ð ‘ †'î ‡œ D£§ ¦ … ¥ ˆ–†… ˆ–†w a@Hˆ–†F ‚ †î ‡ ‡ @V¶ˆˆ–†… ˆ–†…w ¤‡ˆ–†…F ¶ … î ‡ £ † î‡ ‡ … ³ ‘‚½‡ˆ™†a„ Hˆ–†… ‘ @H~–†… ³ ~–†ð ‚ †î¶ '%½‡~–†… ‡~–†…ð ~–†F ³ ‘ ~–†… H~–†ð ‚ ˆ–†F î £‡ … ‡ î‡ … ‡ … † Risk premium €  Ü Ü € ‚ ~–†… ‡ ‡ H‡ … … ² à § à  Ü  Ü 0 á Š Ü â Ü   á § Ü € Š ‡‡ … Œ H~–†… • –†¢ Ü   â   á  €   Þ ¬ à Ÿ Ü Ü Ü Þ ‘º ‚  † º ³  ‘ @î ‚ î  ‘ @î º ‚ Hˆ™†… • –†¢ î † ‡‡ … â Ü â Ü € Š Ü â  Þ Ÿ á Š à  â ! â Ü † Ü Þ ¶ yƒ ³ † • Ü ƒ   2á à à p ‡‡ ¿ˆ–†… • ™†…  ¢ Ü â ‡ … ˆ–†F  Let denote the wealth of the investor at time . Itˆ ’s Formula o 171 can be random. The investor finances his investing by borrowing or lending at interest rate . Equating the -hedging rule: . but must be adapted.

The integral form of the CIR equation is : for every . We obtain . then . where exhibits mean reversion: .7 Mean and variance of the Cox-Ingersoll-Ross process 172 The Cox-Ingersoll-Ross model for interest rates is and are positive constants. This is ‡‡ … Hˆ–†y … î á c ‡ ~–†… î á T ‘ … ‡ … ¢‡ ƒ 3 ‚ ˆ–†y á á If We solve for Integration yields which implies that Differentiation yields Taking expectations and remembering that the expectation of an Itˆ integral is zero. this equation is .¦  á £ ¤ ¢ Ì ‡ … ‚ ˆ–†y ‰— a å á … ‚ ‡ ƒy a å á ¦  á Ê å jb ¦  á å —b ¦  ˆ G f æ å ¬jb ¦ d á  á ä ‰— º å —b ‡ … ˆ–†y ‰— á Œ º ¦ d ì º ‡ … Œ ‚ ˆ–†y ‰— ¥ñò ‡ … ~–†y ‚ … Ž‡ ƒ y † ä £ … ‡ … Œ ܇ ƒ y º ¼ ‘ ‚ ˆ–†y ‰— ä ä … £‡ … Œ‡‹ £ º … ‚ î º ‚ ‡ ƒ y ܈–†y í ä ‡ … @î ‘ ‡ … ä ‡ … † ‚ ¾ ~–†y ‰— î ¢ˆ™†3 ‰— ˜ º ‚ ~–†y å jb á á d å jb á 0 ‰— º å jb ä †@î î !  ‡ … • ˆ–†y ‰— á d  ˆ T £ l ‚ H~–†y ‰— l … ‚ ˆ–†y ‰— 'î† î ‡‡ … £ ‡ … ¦ d á ¦ c d á  ˆ c á ¦ c f d î ‡‡ … £ Œ ÜH y ‰— ¢ … º í  ˆ c ˆ ’ á ã f € ˆ G ‘ ‡ ƒ y ‚ ~–†y ‰— … ‡ … á á ¦ c f d á á ˆ c á  ä á ‰Ž€ á d á  €  ¦ d   á €  d ä á ‰Ž€    á d   á 7¦ d   á 㠀 n á ¦ ’ á 㠀 n  á ¦ c d á    á G á G á c    á G á c … î‡ … ‘ ‡‡ · îÜH† …y ¢ … £ ‘ … ‡ … ¢‡ ƒ 3 ‚ ˆ–†y Œ ‡† a„ ¤ 3 º º í í ‡ … ˆ™†3 ‡ … î ˆ™†a„ ¤‡ˆ–†… ê ‘ î ‡ ¢†'¢ˆ–†… £† ‡ … Ü'î¤~–†y ‡ Ü … ‘ â â ⠆ ‡ … 'î¤~–†y ý~†–a„ ¤ˆ–†… ê ‘‡ … î‡ ‘¢'¢‡ˆ–†… †î £@¤ˆ–†y †î‡ … â â ⠇ … ‘ î ‡‡ … ¢†'¤H~–†y £´ … ˜ ‘ ¾ ~–†…a„ ¤ˆ™†…3 ‡ î‡ ‘ † ‡‡ ¢'îÜH~–†…y £¢ … ˜ ~–†y ‡ … ¾ ~†–a„ ˆ–†…y ⠇ … î ‡ … ‡‡ ~–†y ¤~–†y ˆ–†…y … þf‘ ‡~–†y ˆ–†y … … ‡‡ … ‡‡ … Hˆ™†3 … î á   ‚ ‚ ‚ ‚ ‡ ‚ ˆ™†… î á T   ˆˆ ™@U á ˆ ’ ‰U U   € ‚ h… ‡€ ‡ … î‡ … Œ † a„ ¤ 3 ˆ c U ˆ ’ á ã º í € á f U ‘ î ‡‡ … £ · ÜH† y ¢ … ˆ G ˆ c á ¦ c  㠀 uŠ á º í ¦ ’ d f d ‡ … î‡ … • ˆ™†a„ ¤ˆ–†y ‘ †  … £ @HHˆ™†3 l … ‚ ~–†y î ‡ … á G  We apply Itˆ ’s formula to compute o ‡ ƒy … á • • • € ¦ d  where 15. then á a å á . we obtain o The mean of . If . In integral form.

the vector increments     and integrating the last equation.8 Multidimensional Brownian Motion d € 7¦ á     ⠂ ‡ … ‰— ‚ ˆ™†3 á á   { ¢• z d € 7¦   ‡ ⠂ ˆ™†… ‰— ä †@î î which implies that †@î ‰— î á ‡ ‚ ~–†… Differentiation yields á   ‰— á   Taking expectations. † . then the processes and ‡ ˆ–†… „ y ‡ ˆ™†… „ For each . we have a filtration ‚ ‡‡ • ŒŒŒ ‡ „ ‡ … H~–†… „ y%Q%• ~–†… þ h… ‚ ~–†a„ H î d € d Ì €   d € ‘º ¼ä å jb   d ¦  á Ê á ä ¼   Ì ¤ Ì €   ¦    ¦  ‘º ‘  åjb   d € £ … ܇ ƒ y ‘ ‡ … … £‡ ¿‡ˆ–†y ‰— A¤ˆ–†… £ … ‡ ƒ y ‘   á Ê á Ê ä ¦ â   £ … ܇ ƒ y ‘      ¦ ‘   á   ‰— º á å jb   º í ‡   € ‘ …Tý‡ ƒ … ‚ ~–†… ‘ ‡ ⠇ … ~–†y ¦ d á   á with the following properties: î 15. Itˆ ’s Formula o . š ‡ … ˆ–†y „  C If . is such that 173 v ç . we obtain Each is a one-dimensional Brownian motion.2 ( -dimensional Brownian Motion) A -dimensional Brownian Motion is a process Ì Œº Œº åjb   å —b   ¼ Ì ä ‡ y £ ⠃… á §  ¦ Ê á   ¦ Ê   ¥ ¼ ä ‡ y £ ⠃… º ¼ å —b Associated with a -dimensional Brownian motion. after considerable -measurable. The integral form of the equation derived earlier for º í ⠇ ~–†… î á T º í ⠏ £ ¤† y î‡ … ˆ G ˆ c á Using the formula already derived for algebra we obtain Definition 15. the random vector ‡ ˆ™†… ‡ … ˆ–†m„ î † For each y  T are independent of   . are independent.y € y €  ‡ ~–†… ‡ þ ¼ –†…m„u¤‡ –†…m„ •ŒQ%ŒQ• ‡ˆ–†…mu܇ þ –†a„ £ „ £ … Œ  † %QT þ † ŒŒŒ is  ö ˆ–†… ô ‡ á € Š   á   ¦ d å —b   å —b   ‡ ¾ ˆ–†…   á d ‡ … Œ ~–†y ‡ † @î ‘ ‡ ‰— ‰— î ¢ˆ™†… d ‡ … ˆ–†y ‰— ‘  …º ä ‚ â ä ˆ–†… ‚‡ ‰— ⠘ º á  € Š   ¦ d ‡ • ˆ–†…   á  á £‡ … ‰— â yˆ–†y ‰— ‡ ˆ G ˆ c á ‘  … ‚ ˆ–†… ‡ â   € Š   ¦ d î‡ Œ ¤ … ˆ T ˆ ’   á f d f î‡ … ‰— º í â £ Ü y ‰— º í ‡  € ‘ Ü … ¢‡ ƒ … ‘ â   f ‡ … î‡ ê Œ † a„ ¤ … ˆ c ˆ c @á ä f ‘ î‡  ¤† … ˆ G ˆ c   á f d   Variance of CHAPTER 15.

we have: Theorem 9. so converges to the constant f f £  f { • z £ ’ are independent of one another. Therefore.9 Cross-variations of Brownian motions 174 Because each component „ Proof: Let of and on However. we have the informal equation š š š 15. define the sample cross variation . and each has . First note that be a partition of . ”  ” è è è  ”  ” ƒ ™ ’  P   f –  — ” h y  ” ’      f – — h   f y è { • z  P  P       ƒ T‚ ƒ ‡ … ƒ ‰— ‡ † þ ‡ † œ e£ þ ¹ ™†… þ ‚ ‡ … Œ ŽŒ ‚ C X£ ¹ –†… þ ¼ ¼ ‡ † œ e£ þ ¹ –†… ¶‡ „ ½Q –†… X£ ‡ þ ¹ –†… „ ³ ¶‡ „ £ ½œ –†… s܇ þ ¹ –†… „ ³ ¶‡ „ £ þ ¶‡ „ £ þ Œ ½Q –†… X¤‡ ¹ –†… „ ³ ½œ –†… u¤‡ ¹ –†… „ ³ þ ¼ ‰— ‚   y   y       š š f – — h š y   f è ‰— ‚ ‡ f è … { ¢• z   P y   y  P  P ‘ y ‘ y ‘ ‘ š š š š  P ¶ ½‡% –†… „X£¤‡ þ ¹ †™… „ ³ ½‡% –†… X£ ‡ þ ¹ –†… „ ³ Œ ˆ‡ –†… u¤‡ þ ¹ –†… „ ³ ˆ‡ –†… X¤‡ þ ¶ „ ¶ „ £ ¶ „ £ ‡ „ £ ‡ „ £ ¾ œ –†… u¤‡ þ ¹ –†… „ ˜ ¾ œ –†… s܇ þ y   y       š  P f è Œƒ yT‚ f è ‰— 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. .49 If We compute ‡ … f ¹ –†… „ ³ þ ‘ ¼ â ¹ –†… „ ˜ þ ‚ ¼ ‘ h y – j h š y   { • z y ’ But expectation All the increments appearing in the sum of cross terms are independent of one another and have mean zero. Therefore. For v ç ƒ i y š ‚ l † ‡ î ‡ Œ 'î ‚ ˆ–†… „ ´~–†… „ î is a one-dimensional Brownian motion. we have . . It follows that   y  P y    P š & f è „ „ š As . and .   y  P y     f è þ ¶ ¶‡ „ £ þ Œ ½‡œ –†… „X£¤‡ ¹ –†… „ ³ ½Q –†… X¤‡ ¹ –†… „ ³ þ ¼ ‚ š š y ƒ i f ¶ ö •ŒŒŒ  † †%Q%• to be y € v •yƒ ³ † e‚ ô ç Ž ¶ yƒ ³ ‚ • ƒ ‡ î ‡ T‚ ˆ–†… „ ´~–†… „ î .

we write ‡ †… ˆ c y ‡ ‡ ‡ … • †… • l ˆ c ê ˆ c ˆ c     ‡ f f š ˆ c   ˆ c ˆ c   ‡ f ˆ T ˆ c ê Çf é ‡ „î‡ ‡ þ „î‡ þ ‘ î‡ ‘ … ‡ … Œ † … ‚¤† … º ‘ † … ‚¤† … º ” ¤ … º ý‡ ƒ c ‚ ~–†ð í í í ‡ „î‡ þ ‘‡ ý … þ „ ¤ … Hþ î‡ þ ‘ î ‡ … · ´† – ‡ ƒ F ‚ ˆ–†F ‘ … ‡ … •  … ‚¤† … é ˆ ’   ˆ c   ‡ º í ˆ ’ ˆ c ‡ º í f ˆ T ˆ c 3 º í f Let In integral form. The formula generalizes to any number of processes driven by a Brownian motion of any number (not necessarily the same number) of dimensions. the quadratic and cross variations are: . are called semimartingales. Itˆ ’s Formula o and be processes of the form and é ‚ ‡ ¨• € • –†… î ß U U ‡ ¨• € • –†… ß é @ î   î é @ î ÿî é @ î  ¢ î   é  where U U . with is Let be a function of three variables.10 Multi-dimensional Itˆ formula o To keep the notation as simple as possible. Such processes. plus a Riemann integral. this equation . . and ˆ ’ ˆ c y ‡ y ‡ p v ç ˆ c é be semimartingales. 175 . we write the Itˆ formula for two processes driven by a o two-dimensional Brownian motion. and let have the corresponding Itˆ formula: o Given these two semimartingales CHAPTER 15. Then we . The adaptedness of the integrands guarantees that and are also adapted. plus one or more Itˆ integrals. for as decribed earlier and with all the variables filled in. The integrands o and can be any adapted processes. In differential notation.š š   F š     ‡ Š š   ‡ f F   ê ‡ Š ‡ Uf U ˆ T î Ó¶ F iF U ‡     ‡     ‡   c   F U U     ‡   ‡ Š   ‡ ‡ ’   ‡      ‡ c   F U U 3 f ‡ … „ ‚ „ ‡  … ‚ ö •'‚e • ‡ †… ⠋ô ‘  þ ³ ‘ þ „î ¶ þ • „‚î·¶ º í þ þ þ ‘ þ … þf‘  ‡ þ ‘ þ Hþ … ‘ ’ ‡ þ ‘ Hþ … f‘  ‡ … … • H‡ ƒ " • ‡ ƒ F ƒ … U U é  U ‡ … • F• … ‚ þ ‘  Hþ ³ ‘ º í ‘  ‘º ³º í ‚ £ ‡‡ … ‡ … ÜHˆ™†Â • ˆ–†F • –†… ˆ c  ˆ c U é U   U ¶ Œ·î î é ‰ é ‰ F F U ‘ î é ‰  ¢ F       é ‰ F  ¢ ‘ † a@î º ‚ ‚ ‚ ‚ ‚ U U U U ‡ „î       ‡    ‡ Š ê     H f       ‡ )     º þ © „‚î D¦ „ î ¥ þ ‘ © „ î ¨¦ þ „ î ¥ þ Hþ § § „ þ ‘ • ‡ ‚î   ‡       ‡ Š     i    ‡ Š   ‡  î  ‘ î î ’ ³ þ ‘ î ‘ î   ⠇ ~–†…𠇈™†… þ †@ þ ‘ þ ¿þ … þ ‘ þ „ î þ ӆ@î %‡ ‚î þ ‘ þ „ î Hþ ¢'w … ‘ … „ ‘†î þ † • @î ‡ ‘ … ‡ „î ‘ þ ‚î þ ‘ @î … „ † þ † þ ¿þ • @î ‡ ‘ … º þ § ‘ þ © „ î ¨¦ þ „ î ¥ Hþ â þ „ î Hþ ¢'w … þ ‘†î and and é      ‡   ‡    ‡ ‡ c 3 c   ‡ Š ‡        ‡ Š   ‡    Š   ‡   ‡   c ê       ‡ Š ‡ c   H   ‡ ‡    3 c é  ê é ‰ ‡ ‚ ‚   ‘ þ „ î þ ¢'î ‘† î Œ „î þ †î ‚ „ þ ‘ þ „ î ¿þ ‘ @e ‚ î • ‚î   ‡ Š ‡  3  ¢ 3 é  15. consisting of a nonrandom initial condition.

176 .

1 Stochastic Differential Equations Consider the stochastic differential equation: Here and are given functions. A solution to (SDE) with the initial condition satisfying f  is a process º ö‡ … y º ¨ˆ–†F ô ‡ †€ • –†…  ß £ €  ë ‡ € • –†…  ‡ … „ î ‡‡ … … ‘ †î ‡‡ … ‡ … Œ ˆ–†m‚ý¿ˆ–†F • –†´ ¢'ÜHˆ™† • –†… ‚ ˆ–†F î  (SDE) and Lips- ‡ D• –†¤ Ü€ • –†¤ ß … £‡ … €   € €   € Š í ” • ß £ €    ë  ë d d   í ” ‡ ßD• †–… £†€ • –†… ‡ € ‡ … € • –†¢ í ì ” c f d   €  ¢ d   ‡ †€ • ™†… ‡ †€ • –†… ß D• € • † Ð À ÊÏ d ‡ … ˆ–†F  . there is a constant such that for all f . The solution process will be adapted to the filtration generated by the Brownian motion.e. If you know the path of the Brownian motion up to time .Chapter 16 Markov processes and the Kolmogorov equations 16.. then you can evaluate .1 (Drifted Brownian motion) Let ð 2­ be a constant and ­ Ç ­ î . so 177 Ñ À ýÐ Ê Ï Å ª ð Ñ ª If is given and we start with the initial condition í ì ” c  f † “ Wl† Ò ˆÊ ×Í V ýˆÊ Ï ÐÏ ØÊ ÅÐ Å  † º ¨ˆ†™… ô º ö‡ … y º ¨~–†F ô ö‡ º º ‡ … î ‡‡ … … ‘ ‡ … ’ ‘ … ý‡ †™ ‚ ~–†F ‡ … • Q7’a„ ¿%’7F • 7’¤ ´’ îý‡HQ’7F • h… ºí ºí … • € ‚ ‡ –†F f ì ’ f Let be given. usually assumed to be continuous in chitz continuous in .i. Í ï î Example 16.

2 (Geometric Brownian motion) Let and ï ­ Ç À ýÐ Ê Ï Å ª ØÊ Ð 3V ˆÊ Ï ­ ­ g ñ ª î ª ØÊ Ð 3V ˆÊ Ï Å Ï yˆÊ Ï ÐˆÊ Ï Ð ÅÐ Ê Ê À ýˆÊ Ï ÐÏ Ï ˆÊ ×»Íý ÅÐ ð ð ð ï ñ ð 2­ ÅÐ ýˆÊ Ï ª To compute the differential w.r.t. Consider Ò wË Ê Ê ª Ñ Ò ˆÊ ×Í V ÐÏ ØÊ Ð Ð Ï ÐÏ ®Ð Ê ·Í ˆÊ ·»ÍÏ ª Î ÅÐ ýˆÊ Ï Ï Ê ·Í Ê Ê Ø Ð Ê Ê Ï Ø À ÜˆÊ Ï ÅÐ ª Î ï õ ô ó ò 2ö'u ­ n ‡ ߅ š î ð ·­ Î ð ñ ð ð ƒ  . . if you observe the path of the driving Brownian motion from time 0 to time . Now let x ‰ … Œ ‚ ‡ ƒF € ‚ ‡ ™† … ‡ … H‡ þ –†F … š  º ‰—  ¬ … í ” ‡ … H‡ þ –†F … š ‡H‡ þ –†…F … š ‡ –†F … þW † †   ¬ f f  ¸ Let be given and let be a function. the only relevant information is the value of .r. to compute the differential w.178 then Given the initial condition 16. given that . f † ‡ –†F … † ‡ … ‡ … H‡ þ ™† … š — › ›› ‡ þ … ‡ þ … ­ Œ H‡ –†F … š ˆ pº« º ‰— ‚ ¾ ‡ –†… › H‡ –†F … –š˜ f  ¬  p ø í ” ú… í ” f ø   f f   s Ü ‰—  ¬ the expectation of condition . Denote by be given.2 Markov Property We have the Markov property  ¬ ù 7f … In other words. and based on this information. you want to estimate . treat and as constants: be constants. and compute the expected value of . treat ð c Õ ï Õ ½ ñ ¬ ð 2­ and as constants: Ò Ð Ê ÊÏ Ð ÷ j ª Î Õ Ò ˆÊ ×Í ˆÊ Ï ÐÏ Ð Ï Ð Ø†ÐˆÊ ·Í ˆÊ Ï V ØÊ Ï Ð Ê ×Í Ï y½Ð Ê ·Í ØÐ Ï ï Õ ½ the solution is Î Ò ˆÊ ·Í ˆÊ Ï ÐÏ Ð ­ n Î ñ ¤ ð Ñ ï Example 16. You imagine starting the at time at value . .t. and start with initial Ê Ð V ˆÊ Ï ­ Ž ð Õ ï Õ ½ ­ Ž ª ð ï ­ n ­ g ª Again.

i f s v jG ~–dv ƒ q … ƒ F„‚€i „ S z The derivative is The random variable with initial condition Q q € I §RtPG y i ƒ @D S z '@D and we are making the change of variable Example 16. This is always the case when Q h xv s eG I I . i. ƒ i € "ƒ i S z @D ƒ @D and for every . and variance 179 . In other words. is normal with mean . conditioned on or equivalently. is Geometric Brownian motion: only through their difference has density . I h RI !6 875  þ  or equivalently. Example 16.f f † I Q h ‰v s e¡Q f IG Q h I v s IG r kR‰T©e¡Q f i Ff s v jG Y ~–}wz “ ’ i o c¡ q S VWƒ € z i@D ” c¡wQ 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@y‘iRedHw©©edb a tu Q h P7HsQ s P7b IG b v IG q S Q s IG VTn©PgE q S u‘Q h PHE IG f r Q h I v s IG ¤t©R‰T©P©Q f i f s v jG Y Q Q h IG b v Q s IG b ™VqekRe7p0t©ed–G o “ ’ i •c¡ € Q IG b D Q IG ¡Re7@'RegE i Y I D Q IG E j S Q IG E mXlRPgkuTRPgˆD Q q € IG §RtP(U h I v s R‰T©I s ©I f h Q h gv s eG † I I f Q Q I eeCQ h ‰v s e§dW™˜—–G IG U Y qG v ƒ v ” “ ’  •@©‘Q h ‰v s PG ˆ† I I ‡ „ … Q s IG t©egE S Q ƒ € q ‚ s I € h I TFR0Tt©iReG y q S Q h IG rpiRegE Q h I v s IG U Y tRw©e§u7q f Q IG b D Y I D U S Q IG E ©Redca`XWVTRPHFD A9 ) ¥ & ¦ 0('(¥ ÿ ¦ þ ý €¬ü ¨ ¢ 9 9 9 ¦ @©¦ ¨   ý ©¬ü ¦ (¥ ü   ý ÿ þ ýü ¡'€¬gû 4 ! $% #  B C¦ þ ý €¬ü ü   ©ý % "¡¦ $ #  ¦ þ Yý ) ¥ & ¦ 0('§¥ ÿ £ ¤¢ £  ¦ ¦ §¥ ü   ý ÿ þ ýü ¡€gû ¨ þ ý Y¬ü ! ¨   ý ¡¬ü ü ¦   ý ©¬ü ¦ §¥ ÿ £ ¤¢ Denote by   ý ÿ þ ýü ¡€gû The Markov property says that for 2 2 31 !   " ©  ¥ the density (in the 16.3 (Drifted Brownian motion) Consider the SDE .3 Transition density CHAPTER 16.e. the random variable .. Markov processes and the Kolmogorov equations variable) of ¨ Note that depends on and and don’t depend on .4 (Geometric Brownian motion) Recall that the solution to the SDE Conditioned on .

.4 The Kolmogorov Backward Equation   i ž … ƒ ˆe 6 › ™˜ — • «‘ œ–Œ(wš¡(› wv   i ž  Q IG E ªeRegdS ¨9 9 9 9 CxH‘|8gG –(wš©§› • 6 §  ’ Q ‘ v Q G E œ Œ › ™˜ — h t‹ Ixvd „ › ¥ g# ‡ ˆ† „   S Q ¤G ‘c™g I v ‰p „ I v xd „ S ’ Q ‘ v Q G E  Œ gCxH‘|HG  Ž— t‹ i “ ’ Q h I v s I ‡ ” @© ©R‰T©PG ˆ† Q h I v s I kRxT©PG † „ „ v … ƒ i S S ƒ D Q ƒ € q ‚ s I € h I W@wˆ‡ˆ‡t©ˆRPG v { ƒ D x Q s IG ‰cx'©©egE a tu Therefore. We then write depends on rather than and only through . 180 Consider ­ wý The variables and let and be the transition density. one can compute the expected payoff from a European call: Therefore. Then the Kolmogorov Backward Equation is: in are called the backward variables. where Using the transition density and a fair amount of calculus.° ° 0£ ¦ ü ) ¦ i(¥ ÿ £ ¢ ¶ ü gû ³ £ ³ ³ € ³   ¦ §¥ ÿ ¦ (¥ ­ ÿ £ ¢ ¶ ¶ ü gû £ ° ° 0£ ¦ ü ¬  ¦ (¥ ÿ £ ¢ ¶ ü gû ¶ ° ° ¦ §¥ £ µ¢ £ ¦  ® £ ) ¦ i§¥ ÿ £ |¢ ° ³ £ ° ˆ£ ¦ ÿ þ ý €¬ü   ý ÿ þ ýü ©Y¬gû ³ ³ € ³   ¨ ­ ¦ §¥ ÿ £ ¤¢ £   ý ÿ þ ýü ¡€gû ¨ ý ü ° ° 0£ ¦ ÿ þ ý Y¬ü ´ Žü þ Yý ¬  ¦ ²(¥ ÿ £ ¢   ý ÿ þ ýü ©ižY¬gû ¦ §¥ ÿ £ t¢ ¦ ÿ ý ¬ü ® & ¦ (T¡¦ ý ¬ü € ÿ ý —¬ü & ¦ '©¦ ¬  '¦ ÿ ý j¬ü ý ¬ü ¨ &   ý ÿ þ ýü ¡'€¬gû i@f ¢ B QRIgvpG f  s v Q I v G j wRxa8@dY i f s ‘ Q IG RegE … ƒ ˆe 6 I v ‰a „ ‘ Q IG RegE Q I v  ¢ B Rgd8G f Y Q I v G j qR‰p@pY ’ Q ‘ v Q  G E œ Œ˜  Œ t œ Œ › ™˜ — • exgw|HG –£¤© Ž”‹ e™(wš©§› –S B RPG Q I ¥ f q ¦ 0@D i f › • s “ › # “   „ S q ¦ 7@D i f › • s … ƒ Fe£ ‘ q ‘ q … ƒ ˆe    i ž   qi ž  q œ Œ › ™˜ — • Ÿ(C–(wš¡©–S ’ Q ‘ v ƒ Cxp”–G y h “ d#  ‘ gwv ‡ F† † Q I v  ¢ ŽR‰a8G f ¢ Q I v  i† RxdG f i Ff s i cf s v Q I v G j 0R‰p@ŠY Y Q I v G j qR¡a8@aY ƒ D Q ƒ € q ‚  € I @ˆFR0‡WetPG f ƒ "@D h f † I Q h gv s e¡Q f IG i @f s v jG ¤™Šv ƒ q … ƒ F„e f 16. and (KBE’) ¦  ® ¦ (¥ ÿ £ ‰¢   £ ‰¢ ¯ ´ Žü ý ÿ þ gû ý ü ü gû þ ý   €ý (KBE) ° ±¯ y .   ©ý þ Yý ÿ   ý ÿ þ ýü ¡'€¬gû In the case that and their difference becomes  ¶ ¬ ° þ €ý are functions of alone.

¦ ¡¦ ÿ ¨ À ™ü ü  ! Ž    ¦ 'ˆ£ ÿ ý j¬ü ¦ §¥ ü ! ) i¦ ý ü ® & ¦ §l©¦ ý ¬ü ¨ € ü & ¦ '©¦ ý ü ¨ ü ¬  ¦ ý ¬ü ¨ & ­ wý f  k y f q f … ƒ Fee i f s f i i „ · † Y  y q j iuS º §y f h f † · @CQ f i@f s v jG ™˜v ƒ q v   i “ ’ ” @© · ˆ† y ƒ ‡ „ S Q ƒ € q ‚ · TFR0Tq™G y f Q IG b D Q IG ¡Re7@'RegE Y I D Q IG E j S Q IG E mXlRPgkuTRPgˆD f f · † B f Q · U v q v ƒ e"@H½w—–G Y y · † „ v · Q · U v q v ƒG "@w½H”™§U º y 6 S S  t y s f Y  U y f · · „ ¾S f Y v f Q · U v q v ƒ e¼cp»g”™G y y f q ¹ · · q ¹   t ž S Y S ¢ f ¹ · U v q v @H½w—ƒ · U v q v @p»Hpƒ ¹ y y y y · q ¹  f S S · U v q v @w½H”ƒ ¹ y y y f · † · · † f B Y Y «v 6 S „ Q · U v q v ƒG "@p»Hp–§U Q"·@Ug»qpv”–G v ƒ y   ” “@© · ‡ˆ† ’ · ¹ h · † · † ž v „ v ¢ f CQ"·@UHv»qpv—–G ¹ ƒ feQ"·@Ugv»qpv”–G ƒ   ” “@© ’ · F† h · † · ¹ · ¹ ‡ ž S „ v ¢ S f · eQ¼cUgv»qpv”ƒ™G ¹ ¹ º (y y   ” “@’© · ‡F† „ S Q ƒ € q ‚ · TFR0¸`–G y f h · † f Q Q · U Y qG v ƒ Ce"@aW––”™G v Q IG b D Y I D U S Q IG E Re7@d`XWVTRPHFD Let 16. Therefore. Markov processes and the Kolmogorov equations Consider be a function.5 (Drifted Brownian motion) CHAPTER 16.6 (Geometric Brownian motion) This is the Kolmogorov backward equation. Example 16.1) 181 . and define ¿ (5.5 Connection between stochastic calculus and KBE It is true but very tedious to verify that satisfies the KBE Example 16.

Then . the Kolmogorov backward equation implies 182 .50 Starting at 1 ¨  À ™ü û ¦ 0£ ­ ÿ ý jü Itˆ ’s formula implies o Å Æ  ¿  ¦ ¯  ¦ §¥ ü ÿ ü ¿ ! ¬ ü # ¦ 0£ ­ ÿ ý —¬ü ¦ ˆ£ ¿  ¿ where 2 x1 Let 4 1 5 þ so Proof: According to the Markov property. be an initial condition for the SDE (5. Therefore.1). We simplify notation by writing . the process satisfies the martingale property: (Markov property) rather than  .) ý & i ³ ¿ € ³  ¨ & ­ ¨ ® §& & i ¿ ³ ¿   € & ­ ¸ý ¨ ­ ¬ & ¿ ¿ ­ wý ­ wý &   &  ¿ ¿  ¦ w©¦ ý ¬ü ¨ & ÿ ý jü    A  È ¸R¦ Ç ¦ ˆ©¦ ý ¬ü ¦ ©¦ 9 À9A 9Éü ¦ 9 ©¦ ¨ À ™ü ¨ )i©¦ ¦ Á ÿ Á ü ¨ Žü Ä ! »Ã   ¿ ü ­ ®¬ ¬ ¨ ü A9 9 9 Á Žü 9 B ¦ ý ¬ü B ¦ Á 9A Žü 9 9 9 ©¦ ¦ À ™ü ü !6  6   B ¦ !6 ¨ ÿ ý —¬ü Á Žü 9A 9 ¦ ©¦ ÿ ý ü ¨ ÿ ý j¬ü  ¦ '©¦ ¿ ¨ À ™ü ü ! „»Ã ŽÂ  Ä   ­  B ¦ ý ¬ü 9 9 ©¦ ¦ ¨ À ™ü ü !6 9A ) À 2 ý 2 Á 2 31 ÿ ¦ ©¦ ¦ ©¦ ý ¬ü ¨ Á Žü ¨ ÿ ý —¬ü ÿ Á ¢Žü  B ¦ ¿ 9 9 9 ¦ @©¦ ¨ ý ü Á Žü 4 ÿ ý j¬ü 6 ¿   l¦ ¿ ü  ¦ † (¥ 1  ¥ §& £ t¢ ÿ ý ¯ À ™ü i û ¦ ˆ£ ü ³ € ³   ­ ¦ §¥ ÿ £ t¢ ý ¯ À ™ü û ¦ ˆ£ ü ¬ ­ ¦ (¥ i ¿ ÿ £ k¢ ý ü ¯ ³ €  ¦ 'ˆ£ ÿ ý —¬ü ¦ 0£ ³   ) ¥ & ¦ ˆ§§¥ ÿ £ ¤¢ ý ¯ À ™ü ¯ ý i û ¦ §¥ ü ! ! ü # # # ¯ # !  ¦ 'ˆ£ ÿ ý j¬ü i ¿  ÿ ý j¬ü ¥ & ¦ §§¥ ÿ ÿ £ ¤¢ £ |¢ ¯ ý ¦ §¥ û  ¦ 'ˆ£ ÿ ý j¬ü ¥ & ¦ ('(¥ ÿ ÿ  À ™ü û ¦ §¥ ü À ™ü £ µ¢ ¯ ý ! ¦ §¥ ü  ¦ 'ˆ£  ¦ 'ˆ£ ÿ ý j¬ü ¿ ¿ ¿ 2 À ý ¥ & ¦ §l§¥ ÿ ÿ Àü ™gû Theorem 5.

In fact. one can use Itˆ ’s formula to prove the Feynman-Kac Theorem.51 (Feynman-Kac) Define ÿ öÀ The Black-Scholes equation is a special case of this theorem. This implies that the integrand is zero. 16. Markov processes and the Kolmogorov equations In integral form. and use o the Feynman-Kac Theorem to derive the Kolmogorov backward equation. Remark 16. one based on the Kolmogorov backward equation. as we show in the next section. o Theorem 5.1 (Derivation of KBE) We plunked down the Kolmogorov backward equation without any justification. hence ¿ ³ € must be zero (FK) Ê & ¦ † ©¦ Ê ü ¨ ÿ Ê ŽŽü Ê & i † ¦ ¡¦ ¿ ³ Ê Žü   ¨ ü ­ ³ € ¿ ³ ¬ ­ ­ ¦ ©¦ ¿ Ê Žü  þ ¨ ÿ Ê ŽŽü ) X¦ ÿ Ê Žü ¦ ©¦ ¦ ©¦   À ™ü ® & ¦ (T¡¦ ³ Ê Žü ­ ¨ ü ¨ Ê Žü !   c ŽÂ  ¿ ü ¬ ¬ ¨ ­ ÿ Ê BŽü ­  ¦ ©¦ ¿  ¦ l0£ Ê Žü ¿ ¦ ©¦ ¨ ÿ ý jü Ê Žü ÿ Ê ŽŽü ¿ ¨  ü  €   þ þ ¦ ¡¦ ý ¬ü # # ¨ ÿ 1 ­ ­ ¨ ü Ê Žü ÿ ý —¬ü ¿ Ì  ¦ T¡¦ ¿ ¨ ý ¬ü ÿ ý —¬ü ý ¿ . we have ¿   ¿ ¿ 1 ¦ ©¦ ü 183 Thus by two different arguments. and the other based on Itˆ ’s formula.CHAPTER 16. so the integral for all .6 Black-Scholes ) i¦ ý ¬ü ) ý ŠgÊ Ó ÿ Ò i¦ ý ¯ Ê Žü ¦ ³ €   the solution is ³ ¯ Í ü x­ ÿ £  ¦ ¦ ©¦ ý ¬ü With initial condition Ì ® & §¦ ý ¬ü Ì € ­ ¸ý & ¦ ý ¬ü ý ü Ì Í  @'¦ ® ¯ ¦ Ê Žü ý ¬ü Consider the SDE Ì (& ) ¦ iˆ£ ü !  ¦ T0£ ÿ À öÉü and 1  ¦ 0£ ÿ ý jü i ¿ ¦ ˆ£ ü ³ €   ³ ­ ¦ ˆ£ ÿ ý j¬ü ¿ ¿ ¦ ˆ£ ü ¬ ® ­ ü ¦ ˆ£ € Ñ Ð Ï Î £  7mip'¦ ÿ ý —¬ü  ¿ Then ) i¦ ý ü ® & ¦ §l©¦ ý ¬ü ¨ ü € ­ wý & ¦ '©¦ ý ü ¨ ü ¬  ¦ ý ¬ü where ¨ & i € ³ 2 ý  ) 2 31  1  i Ë We know that is a martingale. we have come to the same conclusion.

× Ø × ¨ € With geometric Brownian motion. and is independent of . for 2 9 ¦ ÿ ¨ ü Ù Ú B ×9 9 9 ¦ Ø ¨ ÿ ü !6  × where -measurable ­ eÄ ß ­ eÄ ß 1 1 ü ü ¦ ® € € Þ kÜ Û Ý Ì ý ¬ü Ñ Ð Ï Î Wqiq¦ Ñ Ð Ï Î Wqiq¦ 2 31 ü ! 3  The independence lemma implies  ¦ ¿ Ø ¨ Ø £ ü ! 3  Ì   '¦ À ™ü Ì  Ì  '¦ Ì  ¦ ý ü À ™ü Ì Ì  ¦ 'ˆ£ ü Ù Now where We thus have . we have independent of .) ¦ i©¦ ý ¬ü Ì ¦ ¨ ÿ ý j¬ü ÿ ý j¬ü ü  ¿ ¿   9A 9 B ¦ ý ¬ü A È ¦ ý ¬ü Ç ˆ¦ Ø d¨ !Å £à  9 9 ©¦ ¦ Ì ü À ™ü !6  ) ¦ iˆ£ ÿ ý j¬ü ) Ò ¦ ý ¯ ¦ À ™ü ³ € ³   ¯ Í ¦ ©¦ ý ¬ü ® ¯ ¦ ® ü € ü x­ À ™ü Ñ qiÚ Ð Ï Î ¦ ý ¬ü Ø d¨ ÿ Þ Ò ¦ ý ¯ ¦ À ™ü ³ € ³   ¯ Í ü Ý tÜ ¦ ©¦ ® ý ¬ü ¯ ¦ € ® ü € Û Ñ qiÎ Ð Ï ­ ³ €  ³ À ™ü ¯ Í ¯ Í ü ¦ ­ ¦ À ™ü ® ý ¬ü Ò ¦ ³ À Ò ¦ ÿ ý ³   ü x­ À ý ) i¦ Ø £ ÿ Ö qÒ ¦ ÿ ý ¯ ¦ À ™ü ³ € ³   ¯ Í ¦ ©¦ ý ¬ü ® ¯ ¦ ® ü € ü x­ À ™ü Ñ mip£ wÔ  Ð Ï Î Õ !  ¦ ¡¦ Ì ü À ™ü Recall the Independence Lemma: If then ! !   c    ¦ l0£ ÿ ý —¬ü Define ¿ where 184 is a function to be specified later. is -measurable. is a -field.

the sum of the for some ¿ 1 & ÿ ý jü ¿ at time if ý Finally.51. This gives us the boundary must be 0. Markov processes and the Kolmogorov equations Because formula. . is a martingale: For . Because of this.51 (Feynman-Kac). then also terms in . By Itˆ ’s o is 185 ¿ . ¿ ­ wý ÿ ý j¬ü This leads us to the equation Along with the above partial differential equation. the tower property implies that . if condition is a martingale. we have the terminal condition ¿  ¦ 'ˆ£ ÿ ý —¬ü i ³ ¿ £ ³ € ³   ­ ¦ ˆ£ ÿ ý j¬ü £ @Í ¿ ­ ¦ ˆ£ ÿ ý j¬ü  ¿ Ì  ¿  € ­  ¦ T¡¦ Ì ý ¬ü Furthermore. À ý ÿ ý ¬ü ÿ ý jü ¿ 9A ) À 2 ý 2 g1 ÿ 9 B C¦ ý ¬ü 9 9 ¦ @©¦ Ì ü À ™ü !6 ™á  This is a special case of Theorem 5. we shall eventually see that the value at time of a contingent claim paying . Therefore. ) i¦ Ì ÿ ý j¬ü ® & ¦ §l©¦ ¦ ý ü ý ¬ü Ì ý ¬ü ¦ ÿ ý j¬ü ý ¬ü Ì cÍ & ¦ '©¦ Ì¿ ý ¬ü & ý ¦ ©¦ ý ¬ü Ì ÿ ý —¬ü ¿ 9 ©¦ ¦ Ì ý ü ÿ ý —¬ü 6 ¿  2 2 2 31 À ý Á ¦ ©¦ Ì À ™ü ü ! ý  ¦ l©¦ ý ¬ü Ì ÿ ý j¬ü We have shown that ¿ CHAPTER 16.) ¦ i0£ ¦ 0£ æ eÄ ç å è ­ ¦ 0£ ÿ ÿ ÿ ý jü ÿ ý jü ÿ ý ü jgÊ ­ ­ i ÿ ý j¬ü ¿  ¿ ¿ £  ¦ ý ¬ü ¦ ˆ£ ÿ ÿ ý j¬ü  Ê ­ i  æ eÄ i'ˆ£ è ç å  ¦ Ê  æ eÄ i'ˆ£ è ç å  ¦ Ê æ eÄ ç @Í ¯ 'ˆ£  ¦ è å ­ ÿ ý j¬ü ÿ ý j¬ü ÿ ý j¬ü ¦ ˆ£ ÿ ý ü j¬nÊ ­ æ eÄ ç 'ˆ£ è å  ¦ Ì ¦ ˆ£ ¦ ¡¦ Ì ü ý ÿ ý —¬ü ­ ­ 1 ¦ ü æ eÄ q@ è ç æ å ÿ ý ü j¬nÊ À ™ü 2 31 !   c  ¿  æ eÄ q@'ˆ£ è ç æ å  ¦ ¦ ©¦ Ì ü À ™ü ! ) À 1  ¦ ) 2 ý ÿ ! 1 Å  '¦ 1 ÿ ý —¬ü Ì È !  ¦ ý ¬ü 1 À ™ü Ó À ©ÿ ä ¡ý ÿ À öÉü Ì £ ÿ ¦ ˆ£ ü  ¦ T0£ ) 1 £ Ó 2 x1 ÿ À C€Wý ã ÿ 1 & ý ¦ † ©¦ Ì ý ¬ü ÿ ý —¬ü i ¦ ¿ Ì ý ü ý ¬ü ³ Ì ³ & € ³   ­ ¦ ©¦ ý ¬ü ¦ ©¦ 9A ) ¦ i©¦ 9 9 9 ¦ @©¦ Ì Á Žü Ì ü Ì ÿ Á Žü  ¿  9A 9  B ¦ 9 Á Žü A9 9 9 9 Á ü 2 71 B ¦ ý ¬ü BC¦ Á 9A Žü 9 9 9 ¡¦ ¦ À ™ü ü !6 ™â   6  B ¦ À ™ü !6 2 ¦ ¡¦ Ì Note that the random variable whose conditional expectation is being computed does not depend on . ÿ ý —¬ü ¿ This is a special case of Theorem 5.

. a put). for geometric Brownian motion (See Example 16.6. Plugging these formulas into the partial differential equation for and cancelling the pearing in every term.g. satisfies the Black-Scholes PDE (BS) derived above.4).1 £ Ó ÿ ò ¦ ÿ ¯ £ ´  ¦ ‘ˆ£ ü ÿ À C™ü ) 1 ö g£ 2 x1 ÿ À C€”ý ã ÿ 1  ¦ ˆ£ ÿ ý —¬ü i ¿ Ì Ì ü ¦ 0£ ü ³ õ ³   ­ ¦ 0£ ÿ ý jü ) ò ¦ ¦ ÿ ý ¬ü ¯ ¦ ´ ® & ¦ §'©¦ À ™ü ý ¬ü  Ž    æ „Ä qæ ²l0£ è ç å  ¦ & ¦ ý ¬ü ­ ü õ „¾¸ý ­ ¦ 0£ ¯ ò §¥ ¦ ÿ ý ü jgÊ ¢ B ¦ ý ¯ À Éü ³ € ³   ¯ ¦ ý À ™ü Í ­ ¯ £ ´ ï îí "e6 ý ´ Žü éÀ ¢ B ¦ ý ¯ À ™ü ³ € ò ©¦ ´ ) ¥ & ¦ 0(§¥ ÿ £ l¢ ¦ (¥ ü À ÿ ýü ¡j¬nû ¦ ¡¦ À ™ü Ì ü B ¦ ð "³ ý ¯ ¦ À ™ü ³ € ³   ¯ Í ü ¯ ¥ £ ï îí ¼„6 ³ € ¦ ý é ¯ À ™ü ê ) 1 £ Ó 2 g1 ÿ À ö€”ý ã ÿ 1  ¦ ˆ£ ÿ ý j¬ü i Ê ³ £ ³ € ³   ­ ¦ ˆ£ ÿ ý j¬ü £ @Í Ê ­ ­ æ „Ä ç å è ¿ In terms of the transition density Even if is some other function (e. we have the “stochastic representation” Compare this with the earlier derivation of the Black-Scholes PDE in Section 15. ¯  ¦ 7§¥ ô € ž ³ ¦ ­   ý ¯ ¥ ü  ¦ §¥ ü ! ñ þ p#  è ç å  ¦ !     æ eÄ qæ Úˆ£ ¯ qiÎ ì Ð Ï The Feynman-Kac Theorem now implies that Ì Í  c²l¦ ý ¬ü ! ­ ­ æ eÄ q@Ú è ç æ å ÿ ý ü j¬nÊ ¦ ý ¯ ë "ê À Éü é … ¥ €  ¦ '§¥ ÿ £ ¢ À ÿ ýü ¡j¬nû ¦ 0£ ÿ ý —¬ü  Ê €­ ¦ 0£ ÿ ý ü —¬gÊ Í ¯ and In the case of a call.7 Black-Scholes with price-dependent volatility ü ! ó ´ ¯ ­ Í  æ eÄ q@å è ç æ £ ­ ´ À ™ü also satisfies the terminal condition £ cÍ ¿ ­ ¦ 0£ ÿ ý jü ¿  ¿ ­ ¦ 0£ ÿ ý jü Í ¯ ¿ ¿ ÿ ý jü Ì §& ¿ . we obtain the Black-Scholes partial differential equation: 186 16. is still given by and (SR) (BS) ap- ¦ §¥ ü ! ¯ ¯ ï îí ¼„6 ý ô T€ éÀ ž ó £  ¦ ‰l0£ ÿ ý ü —¬gÊ .

CHAPTER 16. Markov processes and the Kolmogorov equations
À
ý ÿ ÿ ý jü

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:
ÿ

¢

1

)

ö a£

1

Ó

£

À 㠀”ý

À

2

ý

ÿ

2 31

ò ©¦ 2 31 ´

¯

ÿ

ÿ

£

ü1

1 

T0£ ¦ 1 '¦  i

ÿ ý —¬ü

ÿ À öÉü

¿

÷ ¿£ ³ ³

¿

€

 

³

€

where . The “volatility” sponding Black-Scholes equation is
­ ¿ £ cÍ ­  ¿ ­ ¿ Í ¯ é ø 2 ù¾1

decreases with increasing stock price. The corre-

¦

ý ¬ü

)

® & §|¦

2
ý ¬ü

÷ |Ì

2 31
€

­ ¸ý

& ¦

¦

ý ¬ü

1
ý ¬ü

  

l¦ 1

Ì Í  c²¦

æ ÷ Ì

ý ¬ü

¿ Ì (&

ã

188

Chapter 17

Girsanov’s theorem and the risk-neutral measure
(Please see Oksendal, 4th ed., pp 145–151.) Theorem 0.52 (Girsanov, One-dimensional) Let , be a Brownian motion on a probability space . Let , be the accompanying filtration, and let , be a process adapted to this filtration. For , define
À 2 2
ý

and define a new probability measure by
) ä ¢ r§¦
ÿ

Caveat: This theorem requires a technical condition on the size of . If
ÿ r©ã ¨

everything is OK. We make the following remarks:
    ¦
ý ¬ü

is a matingale. In fact,
³ ­ ¦
ý ¬ü ý ¬ü

189

ý

& ¦

ý ü

ÿ ¦

ý ¬ü

³ ý

³

¯ ¦

ý ¬ü

ý

® & (¤¦

ý ¬ü

® & (¦

ð Ê

ý ¬ü

& ¦

ÿ ¦

Ê Žü

ý ¬ü

³ ý è

³ ý

þ

ì Ð Ï # ³ qiÎ

) i¦

 

® & (¤¦

® & (¤¦

2

ý ¬ü

ý ¬ü 

ÿ ¦

ÿ ¦

2 1

Under

, the process

, is a Brownian motion.

ÿ

h

Ê

& ¦

À

Ê Žü

A

³ ý 

ý

2 x1

þ

# ³

2 ²1  

ÿ

¯ ¦

¦À

ý ¬ü

Ê Žü

2

®

& ¡  ¦

ÿ

¦

® & §¦

ý

ý ¬ü

À ™ü

®

2 ü1 Ê Žü ÿ ý

­ ÂÊ

ÿ 

¤ # ¥¦ 

¦

þ

& ¦

ý ¬ü

#

¯ ” miÚ¤¦ Ð Ï Î 

À

Ê Žü

A

ý

¢ £ü 

ý

þ

#  ¦

¡  

ý ¬ü

¦

ý ¬ü

û ©ÿ ý ý
ÿ

¦

ý ¬ü

ý ¬ü

¯

¯

ý ü

A

ÿ ú öü

þ

ÿ

®

þ 

|¦ 

®

ý ¬ü

ÿ &

À

2

ý

¡  

2 31

ÿ

¦

ý ¬ü

ý ÿ

190
1 À À ê ¨ ý ¦ À ™ü 1
¡  
ü

so 
 

is a probability measure.
¨ À ™ü

in terms of

. Let

denote expectation under
¤

. If
¦
¢

is a random variable, then
) "È ä ¨

Now use Williams’ “standard machine”. and
ä ¢ r§¦ À ™ü
¢ £ü ¡   ¡ ¡  

. The intuition behind the formula
& ¡  ¦ ÿ
¤ #

but since and , this doesn’t really tell us anything useful about we consider subsets of , rather than individual elements of .
À ³ ý  

Distribution of

. If is constant, then
³ ¯ ¦ )i¦ ® ý À Éü ­ €À ® ý ¯ Ñ miÎ Ð Ï À ™ü  ¦ À ™ü À ™ü þ ÿ

Removal of Drift from . The change of measure from To see this, we compute
À ³ ý È ¦ C ¡¦ À ™ü ® À ™ü ­ –À ® ý ¯ Ñ qiÎ Ð Ï ý ¦ À ™ü ÿÅ £à    ¦ À ™ü þ ®   

to

removes the drift from

(Substitute

)

¦

À ™ü þ

®

&

ð ³

­

¯ miÎ ì Ð Ï 

Â¥

) & 

¦ † ©¦

&

ð ³ ¦

ð ³ ¦

$ %À

À ™ü

³ ý

®

À

À

¡  

ý

ý

À

À

 

³

­ €À

ê

ê

¯ 

­

¯

ý

ü

ü

¡

ü

¥ §&

¯ 7mi`¦ ì Ð Ï Î

¯ ì miÎ Ð Ï

Ò

¯ "mi`¦ ! Ð Ï Î

ê ð ³ ¥

³

¯ qi¥ ì Ð Ï Î

À

¯ ¦

ëé "ê

­

­

ô

À

À

ý

ý

ü

ü 

 ¦ &

ñ

ñ

ñ

ñ d#

ñ

ñ

æ

æ

æ

#

#

ä

À

À

À

¦

ëé ¼ê

ëé ¼ê

ëé ¼ê



À ™ü þ

)

ô

ô

ô

1

®

ü

Under , variance :

is normal with mean 0 and variance , so
¡

is normal with mean

À

ý

¡  

¦

Ò

À Éü þ

ú

®

ÿ

¦ 

À

¡ ¦ 

ÿ À C™ü

ÿ 

|¦ 

is that we want to have
1  |¦ 
ü ¡  

) "È

ÿÅ šÆ  

A

¡ &

¤ 

A

À ™ü

# ÿ  ²

& ¡  ¦

À ™ü 

¦ 

¨

To see this, consider first the case
ÿ
¤ ¦ #  ¢ ü ¡  

, where

. We have

ÿ

é 

é

À Éü

ÿ 3  

& ¡  ¦ ÿÅ £Æ 
¡  

À ™ü  

 

#  ÿ ¥¦

é

is a probability measure. Since
ú Žü ®
¡  

, we have

for every

Ó µý 

ý ü

ÿ 3 

ÿ Ô   

ü

ÿ

ú

¦     

À ™ü þ

ý

1

®  

 

­ €À 

¨

¦

¦

ý

À ™ü 

À ™ü þ 

ü 

® À ¡

®

¡ 

ü

¡  

¡  

. In particular

. Thus,

and

.

CHAPTER 17. Girsanov’s theorem and the risk-neutral measure
1

191

We can also see that
þ    2 ® Ñ
¡

Because
³ ¦ ® ý ¯ &qiÚ ! Ð Ï Î ³ ý ÿG%À $ ­ À ™ü þ ³ ¯ ¦ ý  ¯ ¦ ® ý ¯ &qiÚ ! Ð Ï Î ³ ý $%À À À ™ü þ ü ³   ¯ ¦ ® ý ¯ &qiÚ¦ ! Ð Ï Î  ³ ý À ™ü À Éü  
$ %À

we have
Ò À ³ ý   ³ ­ 
ý

Means change, variances don’t. 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:
)
ý

17.1 Conditional expectations under
¨ 2 31

Lemma 1.53 Let

. If

is 
  A

-measurable, then
) È RR¦
ý ¬ü

Proof:
È È qC¦
ý ü

2

2 31

because

, is a martingale under

.

& 

® &) ® & (‚x§

A

Ç ˆ¦

À ™ü

ÿ)

³

¦ ©¦

¨

Å  Å ‰ "à 

È È qC¦

ý ü

( )  '

ÿ)

ý ü

® §&

¨

Å ‰ 

A

Ç ˆ¦

­ ¸ý

¡

À ™ü  È ˆC¦

& |¦ 

È R¦

ÿÅ £‰ 

ý ¬ü

À ™ü

ý ¬ü

¨

¦

ý

ý ü

ü

ÿ)

ÿ) 

¨

¨

¨

Å à 

Å ‰ 

Å ‰ 

þ

® &   

þ

® &

À

¨

À

À

ý

ý

À

Under mean
ý

, is normal with mean zero and variance and variance .

. Under

,

)  &

¦

À ™ü þ

ð

À

®

³ ý

 

³

¡

­

)  & 

ý

¯

¦ ð ³

À ô ë ê ¯ qiÎ À "ê )  & ì Ð Ï é À ð ³  ô ë ê ¯ qiÎ À "ê ì Ð Ï é À ¯  ý ³ ¦ À ü ¯ 7qiVÒ  & Ñ Ð Ï Î ¦ ® Ñ ¡ ä À Éü þ

À ý

À

ê

¯ 

ü

¯ qiÎ ì Ð Ï

À

ëé "ê

ô 

‰Ò  & 

‰Ò  &   

À ™ü þ ÿ ä ä ®

¦

¦ 

 
ý ü

by arguing directly from the density formula

À ™ü þ

® Ñ ¦

À Éü þ

¡  

®

¡  
ÿ

is normal with

¦

ý ¬ü

ÿ

) X¦ ¦ A È ¦ Ç ˆ¦
ý ü

Á Žü ÿ ¦
ý ¬ü

ÿ ¦

Á Žü þ ®Å þ ¡ 

® ¦

¦

Á Žü þ Á Žüé Á Žüé 2 ÿ ÿ

®  

A Ç ˆ¦
ý ¬ü

Á Žü 

È ¸C¦ 2

Á Žü 2 31

®Å þ g  

À
ý

Á    ¦ ÿ ® þ
ü

&
ý

ÿ ý

¯

® §&

& ý ÿ ÿ ­ wý ­ & ® & ÿ r® & þ ­ þ

¯ )g®(&¤¦ ÿ ÿ ý þ® ­ ü ®(& ÿ ý ® ¯ þ & ÿ «® þ ÿ ­

) i¦ ¦
ÿ ý ü ý ¬ü

® & §µ¦ ® (&
¡
ý ¬ü

ÿ ¦

& |¦

ý ¬ü ý ü

ý ý

¯ 

¦  ¦
ý ¬ü ý ¬ü

ÿ & þ ® & ¦ A ÿ ¦ þ ®

­ ¸ý

ý ü

ý ¬ü

) À

2
ý

2 Á

2 x1 ¦
ÿ

® 

È ¸C¦

Ç ˆ¦
ý ü

Á Žü þ

Á Žü

®Å þ ‰  

Although we have proved Lemmas 1.53 and 1.54, we have not proved Girsanov’s Theorem. We will not prove it completely, but here is the beginning of the proof.

Lemma 1.55 Using the notation of Girsanov’s Theorem, we have the martingale property

)

¡   &

¨ ¨ 
¤

¤ # 

    A  & È ¡   ¸C¦ ¨ Ç ˆ¦
ý ü

È È C¦ ¨
ý ¬ü

Å¡  

A

ÈÈ 8C¦ A Á Žü ÈÈ 8C¦ Ç ˆ¦

A ˆ¦ Ç Ç ˆ¦ ÿ ¨
ý ¬ü ý ¬ü

Á Žü Á Žü
ý ¬ü

ÿ ÿ ¦

¨ ¨ Á Žüé A 

¡  ÿ Å ¤  ¡  Ú  Å Å

Å ¡  ¤  Ú  Å ÿ
¤  6

¤

B C¦ È

Å ¡  

  A

¦

È C¦

Ç ˆ¦ A Á ü ¦
ý ¬ü ý ü

A ) È RC¦ 2 À
ý

Á Žü ¦ Á Žüé

Á ü  È ¸C¦

ÿ ¨

Å ‰ 

Ç ˆ¦
ý ¬ü

Á ü 2 Á

ÿ

¨

Å ¡ 

ÿ

A

Ç

Å ‰  

Lemma 1.54 (Baye’s Rule) If

192

is

-measurable and

2 31

¨

Next we use Bayes’ Rule. For
& Á ü ÿ A ä ¨ Å ‰  ¦ Á Žüé

Proof: It is clear that property. For , we have

­ ®¬ Ä
 

¦

0

Á Žü

¢

Therefore,

Proof: We first 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)

ÿ

¤ #

CHAPTER 17. Girsanov’s theorem and the risk-neutral measure

193

Definition 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 defined by and
¡  

of the Girsanov Theorem are equivalent. Recall that
ä ) Þ ¦
5 ¢
ÿ

17.2 Risk-neutral measure
2 2 w1

Stock price:
ý ¬ü

The processes and are adapted to the filtration. 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:
¨ ¦
ý ¬ü

Risk premium
6

©

Market price of risk=

¦

ý ü

@ )9

@

@

@

A

® & §l¦

¦

ý ¬ü

ý ¬ü

® (&

Ì ©¦

­ ¸ý

ý ü ¬S3

&

­ eÄ

¦

ý ¬ü

ý ¬ü

­ wý

Í

¦

ý ü¯ ¬u3

Ý tÜ

& ¦

È

¦

ý

ý ¬ü

ý ¬ü

& ¦

1

Þ Ì ©¦ ¦

Û

ý ¬ü

ý ¬ü

Ì @Í

7

7

7

78

Í

¦

¯ ¦

Ý kÜ ¯ ¦

ý ¬ü

ý ¬ü

Ì ©¦

ý ¬ü

ý ü

Ì (& Å ¦

1

ü

Û

3

¦

¦

ý ¬ü

Capital gains from Stock
5

Interest earnings

ý

Þ

& È ˆC¦

ý ü

Ì ©¦

ý ¬ü

5

Ý kÜ

¯ ¦

ý ¬ü

¨

£

Å ¦

ý ¬ü 

Í

Û

1

­

ý ¬ü

ý ¬ü

ü

5

5

Þ

­ ¸ý

­ ¸ý

­ ¸ý

Ý kÜ Ì & (¦

& l¦

& l¦

& l¦

ý ¬ü

ý ¬ü

ý ¬ü

ý ¬ü

2

¨

¨

¨

5

Û

¦

¦

¦

ý ü

ý ü

ý ü

2 31

Interest rate:

. The process

is adapted. . We can write the wealth process differential in

A

As usual we are given the Brownian motion: , with filtration defined on a probability space . We can then define the following.
ý ¬ü

À

ý

ÿ

¦

ý ¬ü

) i¦

® & §¦

À

2

ý ¬ü

ý

Ì ©¦

2 a1

ý ü

3 4¸ý ­

ÿ

¦

ý ¬ü

& ¦

¦

®

ý ¬ü

ý ¬ü

Í

Ì ©¦

ý ü

¦

1 2¦ 

û ©ÿ

1

A

ÿ ú CŽü

Ì (&

À

¦

ý ü S3

­

è„Ä 0

ý

Í ²

Í ²

Í ²

Ë 

ÿ

¦

¦

ý ü

ý ü

ý ¬ü

1

Í

¨

1

If

, then

)

A

ä ¢

1

ÿ

¡   &

¦

À ™é ü

ÿ

¤ # 

)

1

¢ ü 

¡

¡ &

 

¤

Ë

1

& 

¢ £ü

¡  

If of

, then to obtain

Because

for every , we can invert the definition

¡   

A

ö p¦

& ¡  |¦

À ™ü

ÿ

À Éü ÿ

#  ¦

)

¡   ¢ ü 

& ¡  ”¦

¡

is

À Éü ÿ

¤ 

¢ ü

¡ ¡  

,

equivalent to the market measure . Define The discounted formulas are Notation: Discounted processes: . which makes all discounted asset prices martingales.Definition 17. ¡ Ä ­ ­ ee»T Ã Ä ­­ eeÄÄ ¡   ) i¦ ý ¬ü þ ® '¦ & ý ¬ü Ì ©¦ Ì ©¦ ý ¬ü ¦ ý ü S3 ¦ ¦ ý ü ¬„õ 5  ¢  ¢ ¦ ¦ ¦ ¦ ý ü ¬sõ ¨ ž ž & & ý ¬ü ý üé ¬„õ ý ü ý ü ¬sõ ¦ ÿ ý ¬ü þ ® ¦ & ý ü S3 Ì ý ü ) i¦ ý ¬ü ® & ¦ ­ ÂÊ Ê Žü ý  þ #  ²|¦ ý ü þ ® ) È ‡C¦ ý ¬ü ® (& & ¤¦ ý ¬ü ­ wý ýÅ ¦ ¢ Ì ¡¦ ý ¬ü ¦ ý ü ¬u3 ¦ & ¦ ý ü „õ 5 5 Q  ¢ ¦    ¢ ¦ ¦ ý ü ¬„õ ¦ ¦ ý ü ý ü „õ Ì Ì ©¦ ž ¦ ý ¬ü ý ü ý ü ¬„õ ¨ ž & È R¦ ÿ ý ¬ü ® (& & l¦ & |¦ ý ü ­ ¸ý ý ¬ü ýÅ ¦ Í ¦ ý ü ¬S3 ý ¬ü ý ¬ü ý üé „õ È C¦ ý ü ® & §¦ ý ¬ü Ì ©¦ Ì ¦ ©¡¦ ý ¬ü ¯ ¦ ý ¬ü ý ü 3 ¬u4¸ý ­ 1 Å ü ¦ ¦ Å ý üé „õ ý üé „õ È R¦ ý ¬ü Ì (& & l¦ ý ¬ü Ì ©¦ ý ¬ü ­ Tý Í ¯ Ì ž & ý ¬ü ) ý & ¦ ¦ ý ü ¬„õ Í ¯  ¢ ¦ ý üé ¬sõ ž ý üé ¬„õ & ÿ jý & ¦ ¦ ý ¬ü ý ¬ü ÿ ý ü ¬sõ Í  ¦ ý ü ¬„õ & B §Ë F D D G EÄ ç  ­ å æ ² ¦ ÿ B IË å  F D D  RÄ ç  ²|¦ ­ ý ü „õ ) È C¦ ý ¬ü ¢ ¦ ý ¬ü B Ì D  RÄ ç IË æ å ž F D ­ & ©¦ ý ¬ü 5 Q ¨ & È C¦ ý ü & l¦ ý ¬ü ¨ ¦ Ì ©¦ ý ¬ü ­ ¸ý ­ ¸ý ý ¬ü B Í ¯ Å D  EÄ ç HË æ  å F D B Í ¯ Å D  EÄ ç HË æ  å F D ­ ­ ¢ ¦ ¢ ¦ ¨ ý ¬ü B IË F D D P EÄ ç  ­ æ å ž & & Ì §& & ¦ ý ¬ü ý ¬ü B Ì D G EÄ ç CË æ å ž F D ­ Under Then 194 Changing the measure. TS and are martingales.2 (Risk-neutral measure) A risk-neutral measure (sometimes called a martingale measure) is any probability measure. .

a process . Note that because . whose corresponding wealth process A Risk-neutral valuation. ÿ 195 Look back This is because is a martingale under  ¦ 1 ü ü „õ ¨ . then satisfies 1 ü À € ý ’ h ‘ v e D Q e G iH½g‘(fl` ÿ ¦ ý ¬ü € ’ Q ‘ v Q G `G egH‘|8'a¾S € ’ Q Q G ` |Ce|lbv ¡   € Q IG ©RP'` ¦ ¨ ™ 1 h  # 3dS „ c ™’ Œ s s th r p iqS ‘G –¾S  l¦ 1 ü Y Example 17. i.e.1 Y European call European put Asian call . Consider a contingent claim paying an at time . Girsanov’s theorem and the risk-neutral measure For the market model considered here. 1 we must assume that -measurable random variable W %¦ ý ü ¬u3 ÿ h Ê & ¦ ÿ ­ U  ­ eÄ mÄeç æ ­ 3 eÄ V¤¦ Ê ü ³ ý  þ # ³   ¯ ¦ ý ¬ü ý Ê Žü ® & (¦ Ê Žü ý  þ # ¯ ” qiÚ|¦ Ð Ï Î  Y Y ¨ ý ¬ü where ÿ A ä ¢ ÿ & ¡  |¦ À Éü ÿ ¤ ¦ #  ¢ ü ¡   ­Ä ­ ee•T Ã Ä X  '¦ À ™ü ¨ . À X ¦ À ™ü is the unique risk-neutral measure. so ) B Àü ™sõ 6 X ¦    ) •B  2 ¦ B Àü 6 Ʉõ á   ¦ ¦ X 2 Â1 À ™ü Àü ™„õ ¨ 6     l¦ 5 If there is a hedging portfolio..CHAPTER 17.

196 .

Let .1 We already know that if & ý ¬ü ¦ ý ü & ¦ ý ¬ü ¨ In particular. be a martingale (under ) .e.56 Let be a Brownian motion on the filtration generated by this Brownian motion.1 Martingale Representation Theorem See Oksendal.. be . In the context of Girsanov’s Theorem. Let relative to this filtration. 197 ÿ öÀ ý ÿ ¦ ý ¬ü 2 g1 ¡ ÿ ¦ Ê Žü þ ® l¦ & ® À Ê Žü 2 Ù  ý þ # 2 Â1 ­ ÿ ¦ ¦ 1 ý ¬ü ü Ù Ø  ¦ ý ¬ü Ø ø for some . then ¨ ¦ ý ü 18.11. such that À ý ÿ ¦ ý ü ¦ û ©ÿ ÿ ú )Žü ¦ 2 ¬ 3ü1 ý ÿ ø ¦ ÿ ý ¬ü ¦ Ê ü ¨ ® & §¦ Ê Žü ø  þ # ­ ¦ ÿ CÀ 1 ü ¨ 2  l¦ ý ý ¬ü 2 31 ÿ ¨ ® ¦ ý ¬ü ® ¦ ý ¬ü ¦ ¨ .5. the Brownian motion is the only source of randomness in . p.2 A hedging application 2 2 ‰1   ¡ Homework Problem 4. the paths of are continuous. Then there is an adapted process ÿ .50. ¡ then is a martingale. Suppose that is a Then there is an adapted process . is a process satisfying 2 2 31 A ) À À 2 2 À ý ý A 2 2 Â1 2 31 ý Theorem 1.Chapter 18 Martingale Representation Theorem 18. Theorem 4. suppse that the filtration generated by the Brownian motion (under ). ÿ ¦ ý ¬ü ® & §|¦ ® & (¦ ý ü ý ¬ü ¦ ø ø ý ¬ü  l¦  '¦ ¨ ý ¬ü ý ¬ü ¨ ¨ ¨ Remark 18. such that A Ø ) À 2 ý is -martingale. 4th ed. i. Now we see that if is a martingale adapted to the filtration generated by the Brownian motion .

there is an adapted process 2 71 2 31 ý ÿ B R¦ ý ¬ü 9 9 ¦ Àü 6 ™„õ á   X  ¦ ý ¬ü Ø ¡   ) X 2 À  l¦ ¨ À Éü 2 1 ¦ 5 ¦ 1 ¨ A Let be an -measurable random variable.5. The corresponding wealth process ¨ According to Homework Problem 4..e. such that satisfies . so that ý ÿ ý ü ü ¦ À À Éü X ® & ¦ ¦ Ê ü þ Ê ü „õ ¦ Ì Ê Žü ¦ 5 Ê ü Žu3 Ê ü  þ # ­ ¦ 1 ü ¨  ¦ ¦ ý ü „õ ¨ ý ¬ü ¦ ÿ ý ¬ü þ ® & ¦ ¦ ý ü ¬„õ Ì ¦ ý ü ¦ 3 ý ¬ü 5 u ¢ ¦ ¦ ý ü ¬„õ ¨ ý ¬ü ž 2 & 2 Â1 ý ¬ü Set i. We want to choose and .) X¦ Ê Žü Ù Ú ¦ ¦ Ê Žü ¦ Ê ü Ž„õ ¦ Ì g3 Žü Ê ü Ê 5 ¦ Ê Žü Ù  Ê Žü A9 9 †     ¦  l¦ ü ü ) À 2 À ý 2 ý 2 g1 ¦ ÿ ® l¦ & Ê Žü þ þ # ­ ¦ 1 ü Ø  ¦ ý ¬ü Ø ¦ ÿ ý ¬ü Ù 2 ) À ) À 2 ý 2 g1 ¦ ÿ ¦ ý ¬ü ¦ 5 ý ¬ü ÿ ÿ öÀ ý ) i¦ ý ¬ü þ ® ¦ & ý ¬ü ¦ 3 ¦ ý ü ¬„õ Ì  ¢ ¦ ¦ ý ü ¬„õ Ì ž & ý ¬ü ý ¬ü A ) ä ¯ ¦   Ê ü ¦ ÿ ý ü  ¦  |¦ ý ü ¢ §¦ ÿ ¢ ü ¡   h ÿ Ê & ¦ Ê Žü ³ ý  þ # ³ ® & §|¦ ® Ê ü ý ¡  &¦ ÿ ¤ # À™ü ¯ ” miÎ Ð Ï  þ # ÿ þ ý ® & |¦ ¦ ­ 7Ê ÿ Ê Žü ý ü ¯ 3 ý  þ #  ¦ ý ¬ü ¦ ý ü Í Í  þ ¦ 1 ý ü  |¦ ý ü h ÿ Ê Ì ©¦ ý ü & ¦ Ê Žü # ” Ì ©¦ Ð Ï miÎ  ¦ ý ü ¬„õ ¦ ÿ ý ¬ü ® & §¦ ý ¬ü & ¦ ý ¬ü 1 2l¦  ý ü ý ¬ü 3 4¸ý ­ Ì §& Let Then 198 be a portfolio process. Define the -martingale and choose 5 ­ vb èeÄ T 1 Ø 1 ¨ so that . representing the payoff of a contingent claim at time .

perhaps larger than the one generated by . . . -dimensional adapted process. although it does not tell us how to compute it. Martingale Representation Theorem Theorem 3.57 ( -dimensional Girsanov) dimensional Brownian motion on 2 2 Ú1 F ¦ ÿ y ¡ÿ û ÿ ú )Žü & x For 2 2 31 -dimensional Girsanov Theorem define the accompanying filtration. we have . .3 where so In particular. CHAPTER 18. ¦  |¦ ý ü ¢ ü ¡   h & Ê ÿ ³ ÇÇ 8ˆ¦ & Ê Žü ÿ Ç ý 8Ç  é þ # ³    ‚ ¯ ¦ Ê Žü ¦ ÿ ® & ) §"i¦ € ® ý ¬ü Ê ü ) ¡  &¦ ÿ ¤ # À™ü ¯ ” miÎ Ð Ï ý  þ # ÿ € ® ) ) ˆF) ÿ & ¦ ÿ ­ Ê Ê Žü € ý  þ #  ²¦ ý ¬ü þ & 2 À ý 2 Â1 ÿ ¦ ©¦ ý ¬ü F ý ÿ ) ) Fˆ) ÿ ÿ CÀ ¦ 2 ý ¬ü ý ý ý ü ý ü ® A ÿ öÀ  |¦ ý   1 2ü ýA y y ¦ & ¦ ©¦ ÿ ý ü ® ÿ ) ) ˆˆ) ÿ ¦ ® ü  Š¦ ý ¬ü ® À ý ý ü ž  h Ê & ¦ ©¦ Ê ü ³ ý   ³ ¦ ­ Í ü Ê Žü  þ # ¯ ¦ Ê ü ® & §¦ Ê Žü ý  þ # ¯ ” qiÚ Ð Ï Î ¦ ¦ ý ü „õ ý ¬ü  ¤¦ ÿ ý ¬ü w A9 2 ÿ CÀ ý 2 x1 ÿ 9 9 ¦ B C¦ ý ¬ü A9 9 9 9 ý ¬é ü w    ¦ ý ¬ü B ¦ ý ¬ü ý ¬ü ý ü „õ B ¦ ý ü 9 X ¦ 6 w â  À ™ü ¦ ¦ Àü Ʉõ 6   X ¦ ÿ A 9¦ À 9 ™ü 9 ¦ 9 Àü ™„õ 6 X ÿ    ¦ ¨ ý ü ¬„õ The Martingale Representation Theorem guarantees the existence of a hedging portfolio.a 199 . It also justifies the risk-neutral pricing formula ) X A9 ¦ ÿ  '¦ ¨ À ™ü 9 9 Àü ™„õ X  B ¦ 9 ¦ À ™ü Àü ™sõ 6 X     ¦ ¦ Àü ™„õ ¨ À ™ü A9 ) À ý 2 2 31 ÿ 9 9 B ¦ ý ¬ü 9 ¦ Àü ™„õ 6 X     ¦ ý ¬ü Ø  2 À Ê ¦ ¦ ý ü ¬„õ ¨ ý ¬ü ¦ ÿ Ê Žü With this choice of 2 Â1 5 18.

under ¡   . be a -dimensional Brownian motion on some . Corollary 4. is a martingale under relative to . and let following: A A ® 2 2 À1 F Let ÿ ú )Žü . such that À ý the filtration generated by the Brownian motion ÿ ® ¦ ý ¬ü Ø ý ¬ü ¡ À Ê ü Ê Žü ƒ CÌ ø Ù ¦  ý ü ¬·   ­ ¸ý þ þ ý # # ø & |¦ ü ­ ­ ý ¬ü ¦ ¦  |¦ 1 & ƒ Ì C©¦ 1 ü # ü ý ¬ü 2 ÿ ” ¨ Ø ¦ ©ý¦ ý ¬ü ø Ð Ï Î  qiÚ¦ ý ¬ü ƒ 1 C„¦   ¦  ¦ & 2 €1 ý ¬ü ý ü ® ¦ ÿ ý ¬ü Ø ¨ ) ÿ ) ý) Fˆü ¦ ý ü ý ü „õ Í ÿ CÀ . then there is a A ¨ & . be the filtration generated by . 2 2 g1 F ¦ ƒ Ì C§& ÿ À 2 y ý ¬ü ¦ À ¦ ¡   ý 2   ý ¬ü û ¡ÿ Theorem 4.4 -dimensional Martingale Representation Theorem 2 ý 2 g1 ÿ ¦ ©¦ F ý ü þ ® ÿ ) ) ˆˆ) ÿ ¦ ý ü ž  þ ® ü  ¦ ý ¬ü þ ® ÿ ÿ ÿ ý ¬ü þ ¦A ® ¦ ý ¬ü ý ¬ü & y .200 Then. is a martingale (under -dimensional adpated process ÿ ) relative to . Then we can define the Stocks ‘ ’ÿ Accumulation factor ) h Ê & ¦ Ê Žü Í  þ € ˆƒ ƒ C1 Here. then there is a -dimensional adpated process such that ¦ ©¦ ý ü 18.59 If we have a -dimensional adapted process then we can define and as in Girsanov’s Theorem. If . F ¡   Ù ÿ ÿ ) ) Fˆ) ¦ ©¦ ) ) Fˆ) ý ¬ü ÿ F ¦ ÿ ý ý ü ¬·  ÿ é ) ) ˆF) 2 Ù  ü  ) ÿ ) ¦  ¤¦ À À ý ü ¬ž  2 Ú1 ý ¬ü 2 2 ý ÿ ý ü ý Ù ¦ ý ¬ü 2 x1 2 31  ¤¦ À € ® & ‰§|¦ A ý ¬ü 2 ý ý ý ü ÿ ÿ € ˆƒ 2 31 ¦ ¦ ¦ ©¦ & Ê Žü Ê Žü þ 3 ý ü ÿ   ¦ F ø ® & ) (@i¦ F … ® @i¦ & ) † ‡€ ý ¬ü ÿ ¦ ) ) ˆF) If . and are adpated processes. & ÿ öÀ ý ÿ ¦ ©¦ ý ¬ü ® ÿ ) ) ˆF) ÿ ¦ ý ü ¬ž  ® ü  ¦ ý ¬ü ® 2 ® ü ý 3 ’ÿ 2 à1 2 x1 ¦ û  –¦ ¡ÿ ® ¦ ¦ ÿ ý ¬ü ý ü x 18.58 on ý 21 A ÿ ú )Žü ÿ a -dimensional Brownian motion . the process ÿ öÀ is a -dimensional Brownian motion.5 Multi-dimensional market model .

1 to be satisfied. but there are contingent claims which cannot be hedged. Using in the -dimensional Girsanov Theorem. Martingales and Stochastic integrals in the theory of continuous trading. the market is said to be complete. we need to choose ÿ . The market price of risk is an adapted process satisfying the system of equations (MPR) above. the Martingale Representation Theorem can be used to show that every contingent claim can be hedged. pp 313-316): The risk-neutral measure is unique if and only if every contingent claim can be hedged. Stochastic Proc. Consequently.60 (Fundamental Theorem of Asset Pricing) Part I.CHAPTER 18. Under . There are three cases to consider:  ¡ ý ý ¬ü ý ¬ü ¦ ©¦ ý ü ý ÿ ) ) Fˆ) ÿ ¦ ý ü ¬ž  ý ü  x¦ ý ¬ü ý ) ‘ ’ÿ ) ) Fˆ) F For 5. then it admits no arbitrage. then there is no risk-neutral probability measure and the market admits arbitrage. the discounted wealth process corresponding to any portfolio process is a martingale.): If a market has a risk-neutral probability measure. Case III: (Multiple solutions). Case II: (No solution. (MPR) has a unique solution . Part II.) If (MPR) has no solution. then there are multiple risk-neutral probability measures. (Harrison and Pliska. (Harrison and Pliska.1)   F … † t€ . Finally. A stochastic calculus model of continuous trading: complete markets. pp 215-260. every discounted stock price is a martingale. If (MPR) has multiple solutions. Stochastic Proc. the market is said to be incomplete. The market admits no arbitrage. For Lebesgue-almost every and -almost every . and this implies that the market admits no arbitrage. ¡   ¡   F Market price of risk. so that (MPR) € ® & §|¦ ý ¬ü € ˆƒ 3   F … † ‡€ Þ È C¦ ¦ ¦ ý ü ý ¬ü ý ü ¬„õ € ® §& ¦ ƒ Ì ý ¬ü ­ eÄ &þ ” F • Ý kÜ ý ÿ ­ ¸ý ÿ é ­ ) ) Fˆ) ¦ & ý ¬ü  ¦ ¦  ÿ € ýÅ ¦ ý ¬ü ý ü ¬„õ ¦ ƒ Ì Û & ý ü ¬ž  ý ü Þ ¦ ¡¦ ý ý ¬ü ¦ ¡   ý ¬ü Í ¦ Í ÝtÜ ¯ ¦ ý ¯ ¦ ý ¬ü ý ¬ü ƒ C1 ƒ 1 C„¦  ü Û “   ¦ ¢ ý ü ¦ ¦ ý ý ¬ü € ý ¦ ý ü ¬„õ ƒ Ì ý ¬ü ž € )ƒ & 3 (5. and Applications 15 (1983). Martingale Representation Theorem Discounted stock prices ¦ ý ¬ü 201 Risk Premium € ˆƒ 3 † ‡€ ¦   ý ü ¬„õ ƒ CÌ F … ¦ ý ¬ü Case I: (Unique Solution). we define a unique risk-neutral probability measure . and Applications 11 (1981). Theorem 5.

202 .

2 2 g1 Stocks:   3 ­ ‘©wý We assume 1 In other words. Let . we omit the arguments whenever there is no ambiguity. To g iS g hS F f ¥S f #S g iS g hS f ¥S f #S F F F y y y A ý ¬ü . 2 ¾1  ® ý 2 ÿ öÀ ý 2 31 ¦ ÿ A two-dimensional market model Chapter 19 be a two-dimensional Brownian motion on be the filtration generated by . 3   3 ’‘e– ³ and have instantaneous covariance ³ Accumulation factor: ý ü ¬sõ The market price of risk equations are   ý   3 .Í ¯ ³ 4 ³ ³ 1 ý Í ¯   ) 3 1 4 h Ê & •Í  þ # ” Ð Ï Î  miÚ¤¦ ³ 3  ³ 3 ³ Ì ) & ³ Ì Ì ³  ®§& ® §& ³ Ì  ³ §& Ì Ì §& ý   3t'e–   3     3 ˜–   3 't    & ³ Ì ³  ³ ³ ÿ—ý 3 é ³ Ì ³  ³ §& ³ §& ¯ ³ §& ³ ®§& ³ ¦ ® ®§& ®§& ³ Ì Ì Ì ³ ³ ³ ³ ³ ³ 3 – ü ­ l    3 – & Ì  ®(& ®(& Ì ³ ̧& Ì §& ³    ³  ³ 3 ³   ÿ jý 3       é 2 2 é 1 ö ³ ) ¯ ) B ³ §& ³ ® 3 ö ³ – ¯ é ™ Ql  ­ – ÿ 3 dÿ   3 ® (& ³ È ® §&   3 – ­ ˜—Tý & ³ 1 6 ³  ³ §& Ì Ì &   1Å Ì      Ì §& ÿ A 2 À A û ©ÿ ¦ ý ¦ ÿ ý ¬ü ¦ ©¦ ³ ® ¦ ® ü  »¦ ý ¬ü ® ÿ ú CŽü ÿ öÀ ý ÿ ý ¬ü ÿ ý ü ¬ž  Let In what follows. (MPR) . Note that 203 ³ – ™ ¯ é ­   ý ³ 3 e– . has instantaneous variance has instantaneous variance . all processes can depend on and . but are adapted to simplify notation.

Define be -measurable. Define the . we define . . ¯ é ™ ) •B ³ ® & ³ ³ þ 3 – ­ l  þ ÿ ® & ³ ® & 3 – ­ e—¸ý 3 ­   jsý & 7Í 6 ³  ³ (& Ì Ì  & »Í   Ì    Ì (& †   þ ) i¦ ý ¬ü ³ ® ¦ ® ­ ÂÊ & ­ ÂÊ & ³ ý  ý    þ # þ  ¦ ý ¬ü ³ ® þ þ ® #  ¦ ÿ ý ü ¬ž  ý ü ¬ž  A ) h ÿ ä ¢ r§¦ ÿ Ê & ¦ ³ ý ³ ³ ý ­   ü  þ # ³   ¦ ³ ý ÿ ·  ¯ ³ §& ³ ® ý  ý ü þ # ¯   ® §&   ý  ¡  &|¦ ÿ ¤ # ¦ ¡   À™ü ¢ £ü ¯ ” qiÚ¦ Ð Ï Î  ÿ  þ # ý ¬ü ÿ ¦ @Í ¯   ³ 1 ü ¯ – é 㠖 ‰jã ³ 3 e– ³ … ¯ @Í 3   ¯ 3 ³ 1 ¦ ÿ  ³ ý ü   ž‘3 Í  ¯ 3   1    ý provided é 㠖 ¡kã é ¯ The solution to these equations is 204 Suppose é é ¯ Then 19. but not the variances and covariances.A9 ) À ý 2 2 x1 ÿ 9 9 B ¦ ý ¬ü 9 ¦ Àü Ʉõ 6 X ¡       |¦ ý ü Ø A ¦ À Éü X ) B ³ ® & ³ ³ þ 3 – ¯ ¦ ý é ™ ­   þ ® & ³ 3 ˜– 6 ³ Ì ³ 5 éõ 5 éõ ­   þ & Ì @Í   ¦ ý ® & ¯   3   Ì §&   ¨ Í 5 ü Ì     ¨ 5 éõ 5 éõ    ¢ ¨ & 5 „ & ³ @Í Ì ¯ ³ (& Ì ³ ü ¦ ­ ý & ¯ & ü éõ   ¨õ ž & & ¦ ³ Ì ³ 5 ý ¯ Ì   5 ¯ ¨ Í ü   ­ ³ §& ³ Ì ­   Ì §& We have changed the mean rates of return of the stock prices.1 Hedging when m rqp„4l n o n m Let ¡   is the unique risk-neutral measure. Then (MPR) has a unique solution -martingale .

A two-dimensional market model The Martingale Representation Corollary implies ) ³ ® & ³ Ù þ  þ 205 We have   þ ® & ¢ ³ 3 e– We solve the equations   Ù  ³ 3 ˜– Every A -measurable random variable can be hedged. Assume that  – m s uo 19. Then ) X  l¦ À Éü ¨ 2 2 31 Ø ³ ý ¦ À Éü ¨ we have and in particular. There are two cases: È È     ® §& ® §& ³   3 ­ T4¸ý Í 3 4¸ý ­ ¯ ³ „ 1 & & ³ 1 Å ³  ³ §& Ì Ì é   1Å       ý ý ³   3 Ì  3   Ì §& é ¯ The case is analogous. With this choice of and setting ÿ ³ ® & ³ ³ þ 3 – ¦ ³ ³ Ù #  ³ ÿ ³ Ì ³   ­ l  þ 3 ¯ ü ) ³ ® & ³ Ù þ ³ 5 éõ ³ Ì ³ – é ® & ™ ¯   é Ù ³ Ì ³ ­   3 l't  5 éõ  ™ þ ³ Ì ³ # 5 éõ ­   þ ­ l  ‡  3 Ì   ­ ¦ ® & 5 éõ 5 éõ ž ­ 1   Ì   Ù ü Ø  5 éõ  |¦  ¦ ³ ¢ Ø & Ø ý ü ÿ ¨õ ž   & ü ÿ ¦ ý ¬ü  ¦ ý ¬ü (MPR) . ÿ 5 ¦ Àü ™„õ X 5     ¦ 1 ü Ø  l¦ 1 ü ¨ ÿ öÀ 5 ý 5 for the hedging portfolio . The market price of risk equations are Í ¯   1 „ The process is free.CHAPTER 19. the market is complete. The stocks are perfectly correlated.2 Hedging when  – .

Let or ³ depend on does not appear. if ¦ ³ ® þ ® & 5 éõ ¢ ³ 3 ³ Ì ³ ® (&   5 éõ ­   3   &   ¯   ýÅ Ì   5 éõ ž    ¢ ¨õ ž & ­ È ­ wý & ¦ @Í   3 '’  1 Å ü Ì   Ì     5 éõ 5 éõ ® §& 3 ­   ©¸ý  Í  ¯ 3 1    ý Hedging: Notice that À ™ü X f U and ³ ý Then Suppose is free. . there are infinitely many risk-neutral measures. there is no risk-neutral Case I: measure. Positive depends on be one of them.³ ® ¦ ¡¦ ÿ ³ Ì ¦ ÿ Ì ü À ™ü Àü ™ž  !  X 3   ‘3 A ³ ® X ) È   È   ® §& ³ ® §& &   ýÅ ³ 3 5 éõ   þ ­ wý ü ³ Ì ³ È ­   & ¦ @Í 3 ­ ©¸ý ¯ ³ 1 ³ Ì ³ Å ¡   ³ ÿ   ý ³     ý Í Í ¯ ¯ ³ 2 1   1 2 3 g f g 3 U   '3 ) qæ ç  qæ ç 3 Þ & B B   ® §& & Í ³ ¯ 3³ 1 6 ý ³ Í   ¯ 3³ 1 ® §& ÝtÜ ¯ Û Í  ¯ 3   Í  ¯ 3   1 1 6 6 éõ   éõ ­ ¸ý éõ ¯ B ­ wý & ¢ ¨õ ž & ) ³ Ì ³ é 3 ¯  ³ 5 ÿ   Ì é T3      5 g g 3 U f f 3 U ) qæ ç 5 éõ ö   ¢ g ç qæ È   ® §& ³ & ¦ @Í 3 ­ ©¸ý ¯ ³ 1 ³ Ì ³ Å ü 5 éõ È ­   ® §& ³ ü & ¦ @Í & ý ¯   Ì @Í ¯   3 ­ ‘©¸ý 5 éõ 1 Å ü   Ì §& Ì   ¦ ý & ³ @Í Ì ¯ ³ §& Ì ­ ¦     ü ·  5 éõ ¨õ ž & f There is no solution to (MPR). and consequently. then it can probably not . For example. Indeed ) qæ ç g 3 U W ç mæ f 3 U Case II: 206 have the solution Í ¯ 3³ 1 Let be an -measurable random variable. then there is trouble. The market price of risk equations Set . If be hedged. This market admits arbitrage.

we define the to match Ø .) 1  ³ Ù & Ö TÃ Õ & ) ³ ® & ³ Ù þ ­ l  þ ® &   Ù Ú Ø & ÿ ³ ® & ³ Ù þ  þ # ® &   Ù  ­ l  þ þ # ­ ¦ 1 ü Ø  |¦ ý ¬ü Ø A9 ) À ý 2 2 g1 ÿ 9 9 B C¦ ý ¬ü 9 ¦ Àü 6 ™„õ â   X  |¦ ý ¬ü Ø ¡   To get so We can write CHAPTER 19. we must have -martingale 207 . A two-dimensional market model More precisely.

208 .

Let So the joint density is Then we have: Pricing Exotic Options Chapter 20 Define Without drift. 209 ­ sý ý  ¤¦ ý ¬ü þ ® ¦ À ™ü t ! ¡ ! ¡ .1 Reflection principle for Brownian motion With drift.¦ ÿ ý ¬ü ® ) 1 ö ‘ ©ã ÿ & ‘ ÿ ‘ h & ð ³ ¦ ¯À ê ‘ ¯À ê ‘ ¯ qiÎ ì Ð Ï ê ü ¦ ô À ¯ ë "ê ô ‘ °ü ê À ê  ¯  & ÿ ‘ & ð ³ ¦ ê ¯ ì miÎ Ð Ï ê {ü |æ z ‡³ ë "ê À é ô c c }‘ ° ° ° ° ³ ‘ & ‘ & £ h (& À £ ð ³ ¯ ÂmiÎ ì Ð Ï ñ # À ëé "ê ¯  $ & ä ¦ À ™ü ® ÿ v‘ & ä 1 ö ‘ ©ã ÿ £ §& ‘ ÿ À ð ³ £ ê { "æ ¯ qiÎ ì Ð Ï ñ ê ‘ z ³ ô   $ # À ë ¼ê é ¯ $ ö »¦ À ™ü ® @! ¡  ã ¦ À ™ü ® ÿ v‘ ö p¦ À ™ü t ) i¦ ý ¬ü ® è Ï@wxhþ ¦ u v y y  À ™ü t 20.

210 2m-b shadow path m b Brownian motion Figure 20. .2: Possible values of ¦ À ™ü ® . M(T)) lies in here b À ™ü ¦ t ÿ Figure 20.1: Reflection Principle for Brownian motion without drift m m=b (B(T).

)  ‘  1 ö ÿ  ‘ ÿ ã  G&  ‘ & $ %À $ ³ ý  & ³   ä ¯  &qi©) ý ! Ð Ï Î ¦ ®  ÿ Ǒ ¦  ð ³ & ä ¦ $ ¯À À ™ü ê    ! ¡ ‘    & ê ! ü À ™ü þ ¡   ô ë ¼ê ¯ ì ÐqÏiÎ À À ¦  ¯  ê ê ‘ ü ¯  miÎ ý ! Ð Ï $ %À ³ ý ³ t ¦ ä À Éü þ ® ÿ v‘    & ä ¦ À ™ü   t ñ æ ) $  & † {  † þ  z ! ¦ ä ®  & ä ¦ À Éü þ ÿ ‘ À ™ü   t ! ¡ ¦   ÿ ‘ ü !  z † # { ñ # † z  ¦ '©¦ ® ÿ ¦ À ™ü þ À ™ü   t ü ! Ô  ñ æ ) $  & † {  † þ  z ä ¦ À ™ü þ ®  ÿ v‘ & ä ¦ À ™ü †   t ! ¡   $ €À ³ ý   ¯ ¦ ³ ¯  &qi3¦  ý ! Ð Ï Î  ÿ ‘ ü ! ¦ ÿ  z † # { ñ    t ü #  z † $ %À ³ ý   ³ ® ý eqiq¡¦ ! Ð Ï Î ¦ ® À ™ü þ À ™ü þ ¦ ©¦ ¦ ® ÿ À ™ü ÿ À™ü ¦ !    !      ¦ '©¦ ® ÿ ¦ À ™ü þ À ™ü   t ü À ™ü þ À ™ü ¦    t ü ! Ô  ÿ v‘ ü ! )  ‘  1 ö ÿ  ‘ ÿ ã  &  ‘ & ¦ ð ³  ¯À ê  ¯ qiÎ ì Ð Ï ‘ ê ü ¦  À ¯ ë "ê  ô ‘ ê ü À ê  $  & ä ¦ À ™ü þ ®  ÿ ‘ & ä ¦ À Éü þ ® ÿ   t ¡   ! ) i¦ A À ™ü þ ® è y y iþ Ï w u  @~Q¦ À ™ü   t A   ) &  ¡  ¦ ÿ ¤ # ¦ ¡   ä §¦ ¢ ÿ À Éü ¢ ü ³ ¦ ® ý ¯ CqiÚ ! Ð Ï Î ³ ý ÿG$%À ­ ¬ü þ ý ³ ¦ ý   ¦ ® ý ¯ CqiÚ ! Ð Ï Î ³ ý $%À ­ À À ³ ­ ¯ ¦ ™ü ü ® ¯ CqiÚ¦ ! Ð Ï Î  ÿ ý ³ ý À ™ü À ™ü $ %À   ¦ ¦ ÿ ý ¬ü ® û ¡ÿ ÿ ú CŽü À ý where 2 2 31 Set  Let  Since ¡   But also. Then is a Brownian motion (without drift). is a Brownian motion (without drift) on . we conclude that be a function of two variables. Define 211 (MPR) . Pricing Exotic Options is arbitrary. Under CHAPTER 20. so .

be given. so the value at time zero of is 1 . The payoff at time ¿ is already the risk-neutral measure. ÿ t Consequently.  Hh ¦ Þ 1 Ì ü Ý kÜ ï î "eí é 3 Ÿ Û † è  „Ä ­ þ™  Þ 1 ¦ Ì ü Ý tÜ ï î ¼„í é 3 ž Û œ è  „Ä ­ þ” S ”  ‚ ´ Ö ´ ò Ö † ¯ ¦ $ 1 ü ® qiq¦ ! Ð Ï Î À ™ü þ 3 ¦ þ ® ! Ð Ï Î eqiq¦ ç å Ì T4  è qæ  Õ › B ˆ ‡ † ˆ è ešCf eÄ ­ þ™ 3 „ ˜ – • ’— ­ Ä þ „  ¯ 1 ü ´ $ … S ™ü À 3 ˆ ‡ † è ˜e£ eÄ „  ò ¦ ­ ¯ ¦ ´ À ™ü Ì Õ 6á 'è q‰  ç æ å  1 ç å  ¦   è qæ T©¦ Ì ü ü Ì ÿ 1 ü ) i¦ Ê ü þ  D ® Ï@w~”|¦ u  y iþ y ¦ ý ¬ü ÿ G$   t 3 ! Ð Ï &qiÎ þ Ì  |¦ ‰ ý ¬ü ) i¦ ý ü ÿ ® ¢ ê ¯ 3 ­ Tý Í 3 ž ý  |¦ ý ¬ü  ý   “ ’  © ¦ ý ¬ü Ì  þ ÿ G$ ® &qiÎ ! Ð Ï þ 3    ‘A @ ý @ @ )9 Þ ¢ ÝkÜ ê ¯ 3 Û Í ³ 3 ž ¦ ­ ý ¬ü ® 7 7 3 6 78 Œ „qiÎ Ð Ï Ì  Ž þ ! Ð Ï &qiÎ 3 ¦ ³ ¯ Í ý ¬ü ¦ Ì ý ¬ü ® Ì  ¦ þ $ ‹ý 3 ¦  ® & §¦ ü 9­ ý ü & '¦ ý ¬ü Ì Í  @¦ 3 ­ ©wý ) † ˆ ‡ † è ˜e£ eÄ ­ … S  ò „ d©¦ ¯ ¦ ´ ¡  Ì ü À ™ü  'è q‰T©¦  ç æ å  ¦ ) X¦ ý ¬ü Ì è  Ï w u @~}¦  y y hþ … S „  ò ˜ƒ©¦ ˆ ‡ † è e˜£ eÄ ÿ ­ ¯ ¦ ´ À Ì ü À ™ü ‚ ©ã ´ Æã Let where À ™ü ‰ ŠÌ To simplify notation. where Because 1 ü 20. 212 is the risk-neutral measure. assume that the option is We compute. ý ¬ü ¿   Ì þ ® Ì ý ¬ü ý ¬ü Ì (& ¡ 1 Ì ü where.2 Up and out European call.

leads to   2 g1 1 We compute : and the analysis is similar.£ & ¥ §§& ð À The standard method for all these integrals is to complete the square in the exponent and then recognize a cumulative normal distribution. x=y x ~ B(T) . 213 . Pricing Exotic Options Figure 20. We carry out the details for the first integral and just  ) £ T§& ð £ (& ð À ³ ý   ³ ¯ À ³ ý   £ ý £ §& ³ ¯ £ ý ¦ ­ ³ ˆ£ À¯ ³ ¯   £ ý ­ ³ ý À ð   À ë"ê é À ô ê ü À¯ ¯ ê  { ô ­ ô ¯ ô ¯ ô  ¯ ì miÎ  Ð Ï ‘ # ´ 1 ü ê ü ê  £ 3 ì Ð Ï miÎ ê  z {  z ç è mæ å Ì è m‰å ç æ ç æ è m‰å ´ ë "ê À é ë "ê À é ë "ê À é # ¦  { ¦ ­ ³ ˆ£ À ð ³ ¯     ³ ý ‘ ê ü £ ý ­ ê ¯ ÂmiÎ  ì Ð Ï z  z { À ³ £ ¯ # 1 £ §& ð À ³ ý £ §& £ 3 ì Ð Ï miÎ ê  # ˆ¦ À ³ £ ³ ¯ ü ç Ì è qæ å À ¯   { ëé "ê £ ý ë"ê é ô ¦ ´ ¯ ¦ ô ¦ ¦ ­ ³ ˆ£ ¯ À¯ £ $ ¯ qiÎ ì Ð Ï ‘ ê ü ¡ ¡9 9 † 9  z † ð 9 ð À ¯ ³   ³ ý   £ ý ³ ¯ À ³ ý   £ ý ³ ¯ £ ý ­ À¯ ê ³ À £ ê ¯ ì miÎ › Ð Ï ¯ qiÎ ì Ð Ï ! Ð Ï Î "mi`¦ 3 1 Ì ü ´ ¯ £ $ # ü  z ç å è qæ  ç è qæ å ¯  1 Ì ü ÿ £ §& ! Ð Ï Î qiq¦ 3 1 Ì ü  ü { # z   { ¦ ­ ³ ˆ£ ê À¯ ¥ ê ü ¥ ê ü ³ ý ¯ qiÎ ì Ð Ï ¦ ­ ³ 0£ ¦ ˆ£ ë"ê À ¯ ¥ À ê £ $ 3 ! Ð Ï Î qiq¦ 1 Ì ü  # z # z  ´ ü ç å  ¦ è qæ w©¦ Ë z   { Ë z 1 ü ¿ £ & ¥ & ) ) (3§eˆF) 2 ¦ ‚ Ì ü )  ‘ ¦ À ™ü   t ÿ ¦ À Éü þ ® We consider only the case ã CHAPTER 20. M(T)) lies in here so  2 31 ‘ ÿ ‚ w©ã ã 2 ¦ ´ 1 Ì ü ã ´ The other case.3: Possible values of y ~ m ~ M(T) ~ b (B(T).

we have to obtain In the first integral we make the change of variable give the result for the other three. The exponent in the first integrand is 214 .) ¦ 1 Ì ü ï î "eí é 3   ‘ ÿ ¦ 1 Ì ü ï î "eí é 3   B ÿ h ê ô À h ¯ 3 ô 3ô Í ¯ h   h   À 3 h ê ô À ¯ 3 ô 3ô À ê ¯ 3 ­ ¦   ‘ ¯À  ‘ ê ü c ó ê ô À ô ¯À  Í ­ c À c ‘ ó6 ê h ¢ Í 3ô 3 ž ê À ­ À ê ô 3 h 3ô Í ó ¯ ó h ¯ ­€À ô c ” Í ¯ Ð Ï qiÎ ­ ¯ Ì ü ­ ¦  3ô À   ­ À ô 3 Í ¯ À ô‘  ê ü ô 3 À ê ô 3 ­ À 3ô Í ­ ¯ ó› ¦ 1 ô ‘ À c ‘ c ê À ¯ À 3ô Í ¯ À ô  c À ó ¯ h ­ À ô 3 Í ó › ô ´ c ‘ ç è mæ å ó› ¦ 1 ü ¯ 1 Ì ü ÿ ê À ¯ À 3ô Í ¯ À Ì  ¦ w©¦ 1 ü   ) h ê ô À ¯ 3 À 3ô Í ¯ À ô  c ó ¯ h ê ô À ¯ 3 ¥ ¤ g À æ 3 3ô Í ¥ 3¤ ¯ ô À c ‘ ó› ¦ 1 ü Ì  ¦ æ ¥ ¤ ž ¯ miÎ ! Ð Ï g ¥ ¤ # æ ¥ 3¤ ¦ ƒæ ¥ ¤ Ÿ  { ) X¦ 1 Ì ü ô  ê ¥§& $ ³ ¥ ê ¢ £ (& ð À ¯ 3 ë ¼ê À é ô £ §& ð "³ À w3 ³ ý   ³ À Í ¯ ¯ £ ž £ ý ­ Àé ê 3 ¯ ÂmiÎ  ì Ð Ï z # ¦  1 Ì ü À ëé ¼ê  ê ¯ { £ 3 ì Ð Ï miÎ # z À ³ £ ¦ 1 À ü ô ë "ê ç Ì è qæ å ô ÿ £ &  ¥ §§& ¢ ÿ ô À ¢ À ¦ ê ¯ Í ¢ À £w3 3 ¢ P£À ¯ £  Â¥ ü ) À ý À Í ³ 3 ­ ³ ³   ­ ¢ ³ ê ¯ 3 ¯ ³ À Í À w3 ¯ ¯ ¯ £ ž £ ü Àé Àé ê ê ê ¯ ¯   £ 3 3 ­ ©uÀ À03 ¦ ý À À ³ ý   £ ý ­ À ³ £ ¯ where ¿ ‚ ´ Putting all four integrals together.

the value of the option at the time if . but a similar albeit longer formula can also be derived for .L) = 0 + v(T.CHAPTER 20. We consider the function ï î í ì Ç ‹˜e£ Eæ This function satisfies the terminal condition ® ç Ô·µ ± È Ø ¬ ñI± ² Ø ® ½ ¬ C« and the boundary conditions × Ù ­ ® ­ eQòP’fC« ² ± ­ ® × ¬ We show that satisfies the Black-Scholes equation ­ ® ã hã « ô Ø ô ó dÎ À ô ã « ¾ Ø jÎ à « Î « ¾ µ ­ ² ± × ¬ ËfÕ¯ ± ­ ¬ Ó°¯ ½ Ù Ò ÑÐ PEj½ » º È ¼ ­ ë ê é è ç ± È µ tedtԗI± ¸ ”ÍÌËÊÉe#·µ ¶ µ ´ ³ È Ç Æ Å Ä Ò ÑÐ PEj½ ´ ¼ Ø × À ¸ ® ­ eQò± ² ´ ¼ À § ± ­ ¬ P£°¯ Ç Æ ¶ ÔÈ eÅ ·µ Ä ® × ¬ fC« ¸ ϋP°u² ¶ ± ­ ¬ ¯ ½ ¬ ¯ Õ¬ § ÛÙ ­ ˜HRÇe#åÄáã âáß Qd± à Å æ Æ Å ä à Þ ² È Ù ÚØ × Íµ ½ § 4ÝØ Ù Ø ® × ¬ fC« Ü qÈ .K) v(t.0) = 0 T Figure 20. ª ¨ j©§ « ½ Ø ® × ¬ ’fC« ± If we let we obtain the classical Black-Scholes formula à Â Ö § Ù Ø ©Ý4Ù Ö ÍÂ Ö ­ ð® ± ± ­ ¬ ¯ ® ­ ¬ tPՒeC« ½ § Ù Ø ©õ©Ù à Â Á ¼ ½  À Ù ½ ½ Á µ × Á ¼ ¼ À À Ù ½ ­ d® À § ­ ¼ Î ½ Á Î ¾ ½ Ü ½ ½ ® Ö ® Î À À ¼ ¼ À µ Ü × Ø ©Ù ¼ ½ ½ ¾ ¾ Ø Ù Ù È Ù Î µ ­ À ± ­ ¬ PÕ¯ × ½ ¼ ¾ ¿µ ½ ¼ ºq¹·£€P°u§‹P°eC« ¸ ¶ µ ´ ³ ± ­ ¬ ¯ ² ± ± ­ ¬ ¯ ® ­ ¬ » If we replace by and replace by in the formula for .x) = (x .4: Initial and boundary conditions. Pricing Exotic Options 215 L v(t. we obtain a formula for . We have actually derived the formula under the assumption .

then the Markov property implies .$" " " Ö " ˜e£ Eæ î í ì Ç $" " " 1 " % ± × &f¬ % 4± 1 % ±$" ¬ " " ­ ± ± ÷ teõ¢ 1 ¬ ¯ ® ÷ °deõ¢ 1 ¬ Õ« ­    5 æ Æ Å Ä 76˜¥Q² ë ê té è ç ԗI± ± È µ ½ ¬ ¯ Õ¬ Ç Æ eÅ Ä ß Q² Þ ¬ " e˜£ Eæ î í ì Ç ­ ë ‹é $ƒtÔ·I± ê è" ç ± È µ " " 1 " ½ ¬ ¯ ¬ Ç Æ Å Ä °Ëe¥a!ß f!ß Q² Þ Þ % ± ¬ ± ± ÷ ¢ × ¬ ¯ ® ÷ ¢ ¬ teÝ3f°eb£× §« ­    à æ Æ (E’eÅ Ä ßÞ $" " " ½ Ö ± % Ë× ¬ " ˜e£ Eæ î í ì Ç ­ ë té dtԗI± ê è ç ± È µ ½ ¬ ¯ Õ¬ Ç Æ Å e#Ä ß Qdteõ¤fՒeb)fC« Þ ² ± ± ÷ ¢ × ¬ ¯ ® ÷ ¢ × ¬ ­    à æ Æ Å (Ee#Ä Ö ± t±  ® h±  ¬ ÷ ¢ ¬ ¯ ® ÷ ¢ × ¬ IÝ'× °Šeb0%« ± t±  »­´ æ ­     à æ Æ Å Ä (Ee#Q² ® ¬ ¯ ® × ¬ « à Æ Å Ä ’× °’fCËe#Q² ï î í ì Ç tee Ræ  ­ ë ê$ t" é è ç Ô·H± ± È µ " " ½ ¬ ¯ Õ¬ Ç Æ eÅ Ä ä È µ —I± ½ $" " " ­ ´ âEàæ  ê â à ß Q² Þ ßÞ ± ¬ ± × % f¬ " ˜e£ Eæ î í ì Ç ­ ë ê é teè ç ± ¬ ¯ °¬ Ç Æ eÅ Ä Ö ± ‹±  ® h±  ¬ ÷ ¢ × ¬ ¯ ® IÝ'fÕ|h±  ¬ ÷ ¢ × ¬ IÝ)f°« »­­  æ    à æ Æ Å Ä (Re#vͱ ²  ¬ % ± ‘&Ë× ¬ " ˜e£ Eæ î í ì Ç ­ ë té dtÔÍò± ê è ç ± È µ ½ ® ­ "Qɱ ² § ® h± ¬ ÷ ¢ × ¬ « Iõ£fCQdt± ² ± $" " "  ® h±  ¬ ÷ ¢ × ¬ ¯ ® Ib'fՒh±  ¬ ÷ ¢ × ¬ Ib£fC« Ö ­ }ͱ ²  ¬ % ± × ‘&f¬ " î í ì Ç #e˜£ Eæ ­ ë ê ‹é è ç ÔÍɱ ± È µ ½ ¬ ¯ °¬ × Ç Æ ËeÅ a!ß Þ Ä Ö ½  ½ ö b÷ © § ¨§ Ü ©± ½ ¬ ¥ P¦¯ ® ½ Ù × Ù ­ ® ± ± ÷ ¢ ¬ ¯ ® ÷ ¢ × ¬ hteõ¤× °’"b£fC« ­    à æ Æ ¡ReÅ Ä Ö § ÿ ² ± × ¬ ¯ þ ­ ý × û ú ù ø df°©e©ü&åiQx÷ ² Proof: First note that Let ­ ö ± ­ ¬ ©ŠÓ°¯ Let is a martingale. Theorem 2. If × Ù ±  ¬ I÷ ® e­   On the other hand. if × ö j̱  ¬ I÷ ¬ ¯ Õ¬ Ç Æ Å Ä e#f!ß Þ But when × Ù ±  ¬ I÷    Suppose ­ Ù × Ù 1 b2Ù In both cases. then . we have so we may write . we have . and choose be given and define the stopping time . Then .61 The process 216 be given.

Ù ½ ½ ö b÷ ÷ ­ QI± ² ç ÔÍò± ± È µ ½ ¬ ¯ ¬ °£ñCt± ² ± ½ ± ± ÷ teÝ¢ ½ § ® ÷ ¬ e£§« ½ ¬ §« ½ Ö ÷ b¢ ½ Ù × Ù ­ ® ± ± × ¬ ¯ ® × itËfՒf¬ ã « udËf¬ ² ± × V ± ± ­ ¬ ¯ ® ­ ¬ « tPՒeCQòP¬ ² ± ­ U Ö ± × f¬ V C D8 ± ¬ ¯ À ± Ë× °CtË× ¬ V à Æ ˜Å QòtËf¬ Ä ² ± ± × U à Æ eÅ ¬ 8 Ä Ö ÷ Ù × Ù ­ ® ± × hËf¬ C H8 ± ± × ¬ ¯ ® × tf°’f¬ ã « ± × ¬ ¯ à Æ ‹ËfÕIÀ ˜Å Q² Ä A tËfՒf§« eÅ Ä 9 ± ± × ¬ ¯ ® × ¬ à Æ 8 ® § Ù Ø ©õ©Ù ­ d® ½ Ù × Ù ­ ® ­ eQ² ã hã « Ø ô ô ó À ô Î ã « ¾ Ø ©Î à « Î « ¾ µ Ö ÷ ¢ b'× Ù 1 õTÙ ­ ® ­ "QIt± ² ± 1 ¬ ¯ °® 1 ¬ ã hã « ‹± 1 ¬ ô ¯ ô ó À ô Î ± t± 1 ¬ ¯ °’® 1 ¬ ã « t± 1 ¬ °¯ ¾ —Î ± t± 1 ¬ ¯ °’® 1 ¬ à « Î ± ‹± 1 8 ± ã hã « ô ¯ ô ó dÎ À ô ã « Š¯ ¾jÎ à « Î « ¾ µ 5 Æ ¬ 4eÅ Ä  G F   (à Ö S ã C H8 Q RP 5 Æ « ¯ ŠIÀ BeÅ Ä  1 8   (à G F I Î ± ã hã « ô ¯ ô ó À ô Î ã « Š¯ ¾ jÎ à « Î « ¾ µ 5 Æ ¬ BeÅ Ä  G 2Î F   (à E 7   à æ Æ ¡ReÅ Ä ± ± ­ ¬ ¯ ® ­ ¬ « tPՒe£§Qd‹eb'f°eb0å%« ² ± ± ÷ ¢ × ¬ ¯ ® ÷ ¢ × ¬ ÷ ¢ Ý)× ­ Ö ã C D8 « ¯ à Æ Å ŠCÀ ˜¥Ä Î × 8 ± ã hã « ô ¯ ô ó À ô Î ã « ©¯ ¾ ©Î à « Î « ¾ µ ¬ à Æ Å Ä ’e#ušBtË× °’fCËe#Ä @8 ² A ± ± ¬ ¯ ® × ¬ « à Æ Å 9 Integrate from to ½ For Ù × CHAPTER 20. the Riemann integral A stopped martingale is still a martingale if if . Pricing Exotic Options Because ± ± ÷ ¢ × ¬ ¯ ® ÷ ¢ × ¬ teb'fՒeb)f°« E 7   à æ Æ (E˜Å Ä The Hedge 1 ¬ ¯ Ւ® 1 ¬ §« Ù ­ Let then follows. Then . be the wealth process corresponding to some portfolio ± × Ëf¬ ± × Ëf¬ Then U and We should take . 217 ¬ ¯ Ւ® W X² U ¬ ¯ ® ÷ °eb¢ ¬ « §uŠeb¢ ² ± ÷ ½ ¬ ¾ µ . The PDE is a martingale. Therefore. we compute the differential : is also a martingale.

However. remains bounded. He should cover his short position with the money market. so no money is needed to pay it off. At the boundary. say %. pays off the option. the option has “knocked out”. The value of the portfolio is always sufficient to cover a hedging error of size of the short position. because the stock price has risen. the hedging portfolio can become very negative near the knockout boundary. If the stock does not cross the barrier . Because a large short position is being taken. times the dollar ± § ® × f¬ ã « § where is a “tolerance parameter”. ± § ® × ’f¬ ã § 1.5. a small error in hedging can create a significant effect. is the dollar size of the .218 v(T. short position. and consequently he is left with no money. he covers this short position with funds from the money market. he is now in a region of near zero. x) 0 K L x Figure 20. The new boundary condition guarantees: ´ ± ± ¬ ¯ ® × tË× °’f¬ ã « QdËf¬ ² ± × V ± § ® × ¬ f§« × For but near . The hedger is in an unstable situation. This is more expensive than before.3 A practical issue ± Ø ® ¬ ’× §« × Ü « Y In particular. If the stock moves across the barrier. and is left with zero.5: Practial issue. ½ ½ has the form shown in the bottom part of Fig. 20. x) 0 K L x v(t. 20. Rather than using the boundary condition ® ½ Ù × Ù ­ ® ­ eQò± ² § ® × ¬ fC« § ± ± × ¬ ¯ ® × tËfՒf¬ ã « Q©Ëf¬ ² ± × V solve the PDE with the boundary condition ® ½ Ù × Ù ­ ® ­ euɱ ² § ® ’× ¬ ã « Y § `Î Y 2. He should take a large short position in the stock. Here is a possible resolution.

Ö wÞ  e ® ­ ve4ý Ø ® Ù ½ × Ù ­ Ö 1 8 ± ¬ ½ 1 ® s ± × %i͵ ½ ¬ ü± ô ó À ô µ ¾ ¬ Î ± ± × tf¬ C µ d± ® e &cdË× ¬ ² ± d ® Ø ² ± ¬ dË× °¯ Ö × 8 f°uŠË× ¬ ± × ¬ ¯ ² ± d 8 Ö Ã Â × 8 Ë× °¯ ± ¬ Ç G F ¸ b Ç Æ eÅ ³ Ä Â × 8 ËfÕ¯ ± × ¬ Ç G F ¸ b Ôc² Ö ± × f¬ C D8 ± × ¬ ¯ ËfÕCÀ Î × 8 Ë× °¯ ± ¬ ¾ Define the undiscounted expected payoff ß QòP¬ Þ ² ± ­ U ² ± × ¬ òËfÕ¯ 8 Payoff: a Introduce an auxiliary process ± × f¬ d we have the solutions With the initial conditions Value of the payoff at time zero: Stock: Asian Options Chapter 21 by specifying 219 ® ± it± ½ ¬ d â â ¬ Éb u #ã à ß uÉt® Þ ² ± e Ø ® ’× ¬ 1 à ¬ °¯ Ç ¬ C ©À F Î e cd± ² ½ ½ ¬ ¬ °¯ d q p h f r¤igØ ² ò± .

The differential of the value of a portfolio ± × f¬ V ± Ë× ¬ U ± ­ ¬ PÕ¯ « so that We want to have is (1. Then and the boundary condition 21. where × 1 Ø ô ó dÎ À ô ¾ ã xØ jÎ à 1 1 1 the terminal condition ½ ¬ 1 Start with the stock price 21.1 Feynman-Kac Theorem 220 is satisfies the PDE .1) .2 Constructing the hedge The PDE for One can solve this equation. 1 E ¬ §« ¾ Ø ©Î à 1 E à Å Çæ Æ eHReÅ „òGR® Ä ² ± e Ø ® × ¬ fC« ± 1 ¬ °¯ à G F ® ± × ¬ ¯ ® ×  hf°’Ó@« Ù × Ù ­ ® ± e ¬ b iG€òcòGRe’f¬ ² ± e ® ­ ® × Ø ® ± e ¬ b h€ÉcHGR® ² ± e Î ã hã 1 Ø ® Ù ­ ® ­ eQ² u 'Ø 1 ô is the option value at time .Ö Í 1 8 ± G F 1 ¬ °¯ Ç G F ¸ b c² ®  1 8 ± 1 ¬ Õ¯ Ç ® ± ­ ¬ ¯ iP°® ½ ¸ « bQò± ² ½ ¬ U ® ‚ 31 8 ± 1 ¬ Õ¯ à G F ® ± × ¬ ¯ ® ×  « hf°’#ƒQòËf¬ ² ± × U Ö × 8 ¯ V ¾ µ ò× 8 ¾ U jÎ × 8 ©¯ ± × 8 ± V C D8 ¾ U ©Î À µ U Î ¬ C H8 × 8 ¾ 4Î ¾ ¯ HÀ ¬ Õ¯ ¯ 8 V V V ² ² ² U 8 Ö ± e ¬ G€òb eHRe#QHGRe’f§« à Å Çæ Æ Å Ä ² ± e ® ­ ® × ¬ Ø ® ã « ½ Î Ø Î « « ¾ µ ® ­ "Q² u « ®h€ÉcCt® ± e ¬ b ² ± e Î ó ã hã Ø « À ô ô ô Ö ± e t® Ø ® × ’f¬ ‚ 31 8 Ö wÞ  e v® ½ ® iw Þ  e ® ­ ve©ý ® iw Þ  e ® ­ y"©ý Ø ® Ù ½ × The function One can solve this equation rather than the equation for .

For ÷ Ù × The hedge is given by and boundary condition with terminal condition ¾ µ … Ù ­ ã « QdË× ¬ ² ± where Ü The function ÷ ÷ Ü ©x­ V at time . 1.1) 221 . then (From Eq.Ö Ù ½ ® "÷ × jÜ Ù × Ù ­ ÷ ® A 1 8 ± 1  ¬Õ¯ à t®h±×f¬°¯®’× 9 ® ± ± ¬ ¯ ® × htË× °’f¬ ã ã « … †‡ŠË× ¬ ² ± ‰ˆ V Ö Ù ½ × Ù ­ Ø ® ± ­ ¬ hPòb eIE˜¥QòP’f¬ à Å Çæ Æ Å Ä ² ± ­ ® × … E ® ­ e©ý ® ± ­ iP’® ã hã Ø ® ÷ ¬ « eCQI± ² Ø ® ÷ e£¬ … ® ­ e4ý Ø ® e÷ Ù × Ù ­ ® ­ eQ² Ø … ô ô ó À ô Î ã ¾ Ø jÎ à … … Î … Ö ± ­ ® ± ÷ ¬ ¯ ® ÷ ¬ Pthe£°’"§« eÅ à E  æ Æ Å Ä â à Þ e##ã áß Qò± ² Ø ® × ’f¬ … ± ­ ® ± ÷ ¬ ¯ ® ÷ ¬ PtheՒe£§« ± t± ½ ¬ d ¬ òb E 7 Å Çæ Æ IE’eÅ Ä u #ã â â  ß QCGR® Þ ² ± e Ø ® ÷ ¬ eC« ½  ® Â × 8 ËfÕ¯ ± × ¬ Ç F ¸ b Ôc² Ö ã ± × Ëf¬ « ¯ ŠIÀ ã « u u Î « Š¯ « Î Î × 8 Î C H8 À ± ± ¬ ¯ ® × CH8 ±×f¬°¯bòtË× °’f¬ Î ó ã hã « ¯ À ô ô ô × 8 ± ã hã ¯ 8 ¯ 8 ó « ô Î × 8 ¯ ¾ ±tf°’fC« ± × ¬ ¯ ® × ¬ ² ¾ jÎ à ã « Š¯ « ¬ ñ² ¯ 8 ã « Î × 8 à « Q² ‚ 31 8 ± 1 ¬ Õ¯ à Compare this with because both these processes satisfy the same stochastic differential equation. This value is satisfies the Black-Scholes PDE . We compute . ® ½ Ù × Ù ­ ® Ö ± × Ëf¬ C H8 ± × ¬ ¯ À ± × f°b©Ëf¬ V Î × 8 Ëf¬ ± × U ¾ ² ± × òËf¬ U 8 G F ® ± × ¬ ¯ ® ×  hf°’P„« 8 Take The differential of the value of the option is CHAPTER 21. we compute next the value of a derivative security which pays off . starting from the same initial condition. Asian Options 21.3 Partial average payoff Asian option ‚ 1 8 ± 1 ¬ Õ¯ à G F ® ± × ¬ ¯ ® hf°’×  « QòËf¬ ² ± × U If ± ­ ® ± ­ ¬ ¯ ® ­ ¬ « PthPՒe£§QòP¬ ² ± ­ Now suppose the payoff is a U Ö ± ± × ¬ ¯ ® × tf°’f¬ just as before.

1 While no closed-form for the Asian option price is known.222 Remark 21. ± × Ë—µ ½ ¬ ’ ‘ )À . Asian options. 349–375. Finance 3 (1993). and perpetuities. Geman and M. the Laplace transform (in the variable ) has been computed. Yor. Math. See H. Bessel processes.

× × 22.1 Binomial model. We have no probabilities at this point. Solving the equations. Hedging Portfolio “ 8  Evolution of the value of a portfolio: so that Ö ô  (four equations) 223 ± ¬ ó V ® ± ˜ hԀ¬ ó V ® G V ® G U There are four unknowns: . (See Fig. be given.1) and . 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. 2. we obtain: G U ± ½ ¬ ó ± ô V ® ó ® i±  ®  ó d e± ô ˜ ™¬ ½  ¬ Given a simple European derivative security use a portfolio processes V . we want to start with a nonrandom Ö ± • ¯ • —V µ • –U ¬ ü± ¾ ©Î ´ ¬ ó Î a ó ç •  ® G ® ¯ ó • @V V  ¬ a ²  ² d± ô ® ó ç –U • 1 ” ´ 7Ú² ó  ¬ ô ® P´ U Î Š¾ ö 1 ® ­ e4ý ¾ Let Let be the set of all possible sequences of coin-tosses. The hedging portfolio is a specification of how to do this trading.Chapter 22 Summary of Arbitrage Pricing Theory A simple European derivative security makes a random payment at a time fixed in advance.

because we have a hedge which works on every path. Recall: Define Then ¯ • @V i. ó ç –x U • ç • In continuous time.e. Then ® p p qo r S E ± ½ ® k Ç â n æ ó QRP  ® % £&± ¬ ô j UI ½ ¬ ± ó ¾ ©Î U 8 ± ¾ jÎ ô ´ ¬ ©µ µ • ± Ö ç • 8 ô 1 ¯ 1 ± ¯ ´ ¬ ©µ Ò Ñ PEÐ µ 1 Ò Ñ PEÐ 1 Ò Ñ PEÐ 1 Î S t ¬ €d²  s² ¬ d² ± ˜ ÔR® 1 E Î k m â l æ ® ô ó ± ˜ Ԁ¬ ± QRP  ± ± • ½ ¬ ô ½ ® ¯ ®ó j ó ó Ò Ñ PEÐ UI  U ¯ µ ± ˜ Ö ± ½ ¬ ó ÍòԀ¬ ó ¯ U U ± ¬ µ ±ó ˜ òԀ¬ ½  ô  ¬ ¯ µ ± ˜ —IÔt® ó ¬ ô ¯ ¬ ô 8 8 µ 8 U 8 µ µ ó µ ¾ ©Î ó µ ¾ jÎ ç • 1 1 µ ± ˜ IÔt® ¯ ´ ´ Ò Ñ PEÐ hi fh g©Î ¾ ¬  ô ¬ ¾ ©Î ´ ´ ó ´ ´ U ² ² d± ² ² d± G G ó ó V U   ¬ ¬ ó ó V U . what matters is that the paths in the model include all the possibilities. We reiterate that the probabilities are only introduced as an aid to understanding and computation. and the pricing and hedging formulas on the previous page will give different results. We want to find a description of the paths in the model. They all have the property ô ‚ ó If we change . From a practical point of view. we will have the analogous equation 8 Ö ± • Ö ® • x ¯ • —V • ‚ Ö ¯ • x • • @V ± × ‚ Ëf¬ ¯ µ ± × ¬ ËfÕ¯ µ • –U À µ ó x  ó Ö ¬ ü± ç • x  ¯ ç • • x –U • • ± ¾ ©Î ¾ jÎ ± × Ëf¬ Î ´ ¬ • —V ´ ¬ d² ó ó V ç • x ç • Î ² ² ó • ± × ‚ f¬ ç • • x ƒU • ± × Ëf¬ x ¯ • @V ² µ xU  ó ó ó ç –x U • ç • ² 8 ó À ç –U • • ¯ Ò Ñ ÓRÐ The paths of accumulate quadratic variation at rate per unit time. then we change . Ö “ ô À Q² ô ô ± • ¯ Ò Ñ PEÐ µ ó ç • ¯ Ò Ñ ÓRÐ ¬ ó G Å v w R• u ­ ©ö 1 1 Ò Ñ PEÐ ² bÀ Let .224 The probabilities of the stock price paths are irrelevant..

regardless of the portfolio used. However. Summary of Arbitrage Pricing Theory If we introduce a probability measure under which martingale. Indeed. rather than coin-tossing. These are the paths with volatility . we denote ® % ó $" " " " Ö ô a x % ô ¯ y a ßÞ x ßÞ ² ô y % $ ² ó %$ " " " " ó ó ô ô xU xU ßÞ y ßÞ y ² ² xU G ó G ó xU ² ² G ó U ² U ¾ jÎ ô ´ U Suppose we want to have must have ´ . 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. but that is impossible. • @V • x ƒU 225 is a martingale. we use Brownian motion. Then we Ö S % ‚ • • $ "x "¯ • " " ‚ µ • x • % • G $ "w " Q t"P µ " ó ó ó ó ç • x  ¯ ç • x ç • ç • ¯ ¯ will also be a ßÞ  y I • —V Î • Î a • –x U • ßÞ y zÞ y ² ² % • a $" " " " ó ó ç –U • x ç • ßÞ y û zÞ y . ± × ¬ ËfÕ¯ Ò Ñ PEÐ ô À × ± × ¬ f°¯ Ù × ±×f¬°¯ Ù ­ ® ± × ¬ hËfÕ¯ ô À ­ ·À ö ½ ¾ jÎ • º º Î Ö 1  } • x 8  ´ ó ¯  ó ç • ç • º Î • ¯ x 8  ´ ¬ ©µ µ 1 1 • Ö º ´ q% ² ¯ º ² % x 1 ´ º } º ® ² š% ² Î º 8 } Î • 8 µ 1 $" º " ó " " µ ¾ 4Î } ó 1 ç • x !ß Þ ¯ ç • ´ º ² x} ÿ y ² º  • y û º zÞ } y º ² % . then ¯ | • 7{ –U | • 7{ We need to choose and so that ® ¾ ©Î ´ q² 8  The solution of these equations is Ö ± ¾ jÎ 8 22. the only thing about this probability measure which ultimately matters is the set of paths to which it assigns probability zero. We would like to hedge along every possible path of . and compute º ² „} | • { zÞ y To find the risk-neutral probability measure ½ ˜ ~² • l under which is a martingale. To generate these paths.2 Setting up the continuous model Now the stock price .CHAPTER 22. we need a probability measure. To introduce Brownian motion. ÿ . where is some -measurable random variable. is a continuous function of .

there is an such that  by . wÞ ‚ Since we are interested in hedging along every path. Surprisingly. for . Roughly. For any .226 ± € Š®  . . We want to define ® ÿ ± Ë× ¬ C À so that the paths of accumulate quadratic variation at rate per unit time. The mathematically precise statement is the following: ô ‚ ó ‚ ± Ö ó ½  ® × ’f¬ Ù × Ù ­ Î × ô ‚ ® h± ô µ ± × ©Ëf¬ µ ± ©Ë× ¬ C À  ® × f¬ Î C À À C gû × ó ô ‚ C ‚ 7û Î p h 3if À × p h ¤if ¾ Î ‚ û ± ­ ¬ ¯ P£°Qdf°Ëe¥Ä ² ± × ¬ ¯ à Æ Å × ô ~d± ‚ ² ó p h ¤if ± ­ ¬ ¯ P°udËfÕ¯ ² ± × ¬ ‚ is a continuous function of . the reason for this is the following: Choose . the paths of per unit time. If we replace However. Then. ‚ The most convenient choice of so and is ® ô ó À ô µ ¾ ² i‚ ± × f¬ C D8 ± ¬ ¯ Ë× °IÀ Î × 8 f°¯ ± × ¬ ‚ ¾ ² ± × ¬ df°¯ 8 zÞ is a martingale under . the choice of in this definition is irrelevant. then is a different function. With this choice of . regardless of whether we use or in the definition of paths. except possibly for a set of paths which has probability zero. we will see the same ó ‚  ƒ ó C À  ± × f¬ ® ½ C À Ù × Î × G‚ Ù Î ­ × ‚ G7û Î ± ­ ¬ P£°¯ ® h± p h ¤if ó  ® × ’f¬ Ò Ñ ÓRÐ ± ­ ¬ ¯ P°udËfÕ¯ ² ± × ¬ C À ± ­ ¬ ¯ PÕQdËfÕ¯ ² ± × ¬ ² ± × ¬ dËfÕ¯ ô ó À accumulate quadratic variation at rate $ ®  ¬ ± × f¬ C À Î × G‚ ô Î À × ® × f¬ ó Ò Ñ ÓRÐ ‚ C À Î ×  ƒ × ó ‚ ô  Ù × Ù ­ ± × ’® Ëf¬ wÞ Let  ‚ ½ C . ® ÿ ÿ × × ô ô Ö ÿ ó À ô ó À ô ± Ë× ¬ ± ¬ Ë× °¯ then this set of paths has positive probability when Î p h ¤if × ô 7û ‚ ± ­ ¬ ¯ P°udËfÕ¯ ² ± × ¬ is defined by ± × ¬ f°¯ If a set of stock price paths has a positive probability when ® ÿ ± Ë× ¬ C À is defined by ± × ¬ f°¯ In other words. be a Brownian motion defined on a probability space . the choice of is irrelevant.

Because must be a martingale. We can change to the risk-neutral measure Brownian motion. and then proceed as if had been chosen to be equal to ‚ .3) .3 Risk-neutral pricing and hedging zÞ y Let denote the risk-neutral measure.CHAPTER 22. ® ± hË× ¬ ‰C 8 ± × ¬ ËfÕ¯ ± × Ëf¬ x À Q² ± ‚ Ë× ¬ ± × ¬ f°¯ Then x  8 Ö à Æ Ä h„Šf¬ ² ± × x zÞ y where is a Brownian motion under . ‰C Ö ± Ë× ¬ Ö ï S ± × Ëf¬ ± × ¬ ¯ ËfÕIÀ C H8 E à Eæ ® ÿ Q tP ‰‹ Šˆ Î ± × ¬ ¯ ËfÕCÀ × 8 ˜Å † ä Æ Î I × 8 À Î ± ¬ Ë× °¯ × 8 ËfÕ¯ ± × ¬ ¾ ¾ † ²ò±Ë×f¬Õ¯ 8 ² C ‰C àEæ à Eæ { ê ‰C is a (3. and initial portfolio value so that . Then ® ± × hËf¬ ‰C 8 Evolution of the value of a portfolio: ® ’× 8 ‹ËfÕtËf¬ ± ± × ¬ ¯ ± × V which is equivalent to ± × ‚ f¬ ± × ¬ f°¯ x  Now suppose is a given -measurable random variable. we must have ½ Ù × Ö Ù ½ ­ ® ’i± Ù × ½ ¬ Ù V This is the risk-neutral pricing formula. We want to find the portfolio process .2) (3. ± ½ ¬ a is a martingale under µ ± × Éf¬ Ö ± × Ëf¬ ¬ ¾ jÎ ‰C 8 ± × ¬ f°¯ ± × f¬ ± × ¬ ËfÕ¯ 8 f¬ ± × x ± × 8 f¬ À ± × tf¬ V V V E à Eæ Œ { à Eæ ² ² ± òË× ¬ ² ² ± × Ëf¬ ± × Ëf¬ ± × ‚ f¬ ± × f¬ xU zÞ y U $ 8 xU  ² ɱ 8 ½ ¬ U a ± ­ P£¬ U E so is a martingale under E . We have the following sequence: zÞ y ­ ® ± × % Ëf¬ $" " ± " " ½ E à Eæ Œ Ea à Eæ { ¬ x ßÞ y E Regardless of the portfolio used. If a different choice of ± × f¬ S Q I Î ó RP Î × 8 ±Ë×fÕ¯ ± ¬ À ô ‚ ¬ …ñdËfÕ¯ 8 ² ± × ¬ ô Î p h ¤if × ‚ G„û ± ­ ¬ ¯ PÕQdËfÕ¯ ² ± × ¬ C À U ‡ 22.1) (3. the payoff of a simple European derivative security. Set . ô ó À ô µ zÞ y ¾ has the same paths as . Summary of Arbitrage Pricing Theory ‚ zÞ 227 is made. we have C D8 and is the risk-neutral measure. under which .

3). A probability measure is said to be risk-neutral provided Ö ± × % Ëf¬ $" " ± " " ½ ¬ a x ßÞ y ² ± × Ëf¬ ± × Ëf¬ xU ± × Ëf¬ U 2. such that Ö a and it will end with value y x !ß Þ ± ² ½ ± ½ ¬ a x ± ® £% ½ ¬ ± ® x ½ % ¬ ± ² x Žß Þ ½ % ™± ¬ a ½ x ½ y ¬ $" ² ± ­ ÉÓ¬ ßÞ " ± " " y À ² ± ­ ÉP£¬ ½ ¬ a U ¾ U ½ ¬ x ² ò± U From (3. (3. Ù × Ù ­ ® ± × ’hf¬ U Then (3. the risk-neutral pricing formula is valid provided there is a unique risk-neutral measure. they are adapted to the filtration generated by .1).2 In general. If you start with a C C a Remark 22. . Comparing (3.1) or (3. the value × of the hedging portfolio in 1 satisfies ² ò± ¬ U Ù × Ù ­ ® ± × ’hf¬ V then there is a hedging portfolio .2). which implies that the portfolio process . the risk-neutral pricing formula is still “valid” when and are processes adapted to the filtration generated by . which we want.3) (not by (3. a is given. we see that the hedging portfolio must begin with value .228 1. (3. we first use the tower property to show that zÞ y defined by (3. The “validity” of the risk-neutral pricing formula means: 1. by (3.5) . Define ½ .1 Although we have taken and to be constant.2) (or equivalently.4) implies (3. which we know. is the value of ‰C .1)) is satisfied by the defined in step 2. because we do not yet have ).2). when there are multiple assets and/or multiple Brownian motions. the definition of . which implies (3. If they depend on either or on .4). we decide to define Ö ± × ¬  ± × ËfŠBf¬ ± × ¬ ¯ ËfÕ§À x Remark 22.3) is a martingale ­ ® ± × ’hf¬ U 3. Construct so that (3. We next use the corollary to the Martingale Representation Theorem (Homework Problem 4.2). ½ Ù × Ù ± Ë× ¬ V under . à Ræ Œ { à Ræ ± × f¬ V E ± × ¬  ËfŠ~² ² ± × df¬ ‚ ± × Ëf¬ ± × Ëf¬ xU  V 8 ½ Ù ¬ À × U Ù ­ ® ± × ’hf¬ ¾  ¯ U 2. ½ Ù × Ù ­ ® ± × ’hf¬ V ½ E To carry out step 3.5) to show that ± × f¬ ‰C 8 for some proecss .4) (3. and (3. At each time .

 To see if the risk-neutral measure is unique.1 Assume and variable. Define °  ½—D(®– » ¼ š » • « º F ¸ ” ¨¹’ “ ‘ Example 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. % š  4²• ³ ´" " ›y™—¡'” š š ˜ • – • " " « © ™H§ ¦ ž š š  • – ž  µ›46®Bt6• ‰¸   ¡‰   ¢ © ¦ œ “ “ š  04²• ± Then ° yï ›y¯®¡'” š š ˜ • – • š š ˜ • – • ” ›y™—¡'``’ “ ­ § «ª © § ¦ (¢ £¬H¨6ä ¥ €£¡‰   ¤ ¢ “ š Ÿ ž  0HBt6• œ Example 22. Define  are constant. compute the differential of all discounted asset prices and check if there is more than one way to define so that all these differentials have only terms. We need to identify some state variables. so that % Ë× ¬ ± " ± " is a function of these variables. ‰C 8 ‰C 22. We can take the stock price to be the state .4 Implementation of risk-neutral pricing and hedging To get a computable result from the general risk-neutral pricing formula ® ± % Ë× ¬ ½ a x one uses the Markov property. and $" " $" " ± " " ½ ¬ a x ¬ ß Þ Ëf¬ ± × y ßÞ y x ² ± × f¬ ² ± × Éf¬ ± Ë× ¬ xU U œ ­ (¢ ª ¶ ­ (¢ ª · . . ‘ where ° ½0D(®– » ¼ š » • « ¢ F Ä ¿ ¦¹0y™Ã¾ “ š ˜ • ž ï &y¯Ã™'” š š ˜ • ¾ • ­ § «ª © § ¦ (¢ )´µH¨Âä Á 7¥ €g‰¡   ¤ ¤ ¢ “ š ¿ ž Ÿ ž  À7B'Bt6• œ » ¼ š » • ½0H¡®– “ š  • 046„¾ š  • 46®– Take and to be the state variables.2 Assume and  š š  • – ž  ›4²®4t²• º ¢  ¢ D§ ¦ © “ and is a martingale under are constant. the stock price and possibly other variables.CHAPTER 22.

Ò Ñ ½Ð we can solve for × 8 ± × f¬ V . we get an expression involving to be a martingale.5).4) to obtain (3. 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. This is the argument which uses (3. The term in the differential of the equation is the differential of a martingale.3 (Homework problem 4.2) is a martingale under and 230 and š  • 46„¾ š  • 46®– Then ± is a martingale under and to be the state variables. and since the martingale is « ‰¸   š š  • ¾ ž š  • – ž  &46„›µ46®Bt6• œ Û » ¼ š š » • ¾ ž » • )½0›D(Ì&'¡½ ¢ º F Ó “ Ú š  4²• š  46• Å ± ³" " " ¢ F « " š›šy˜¯•®–¡•'”ÈÔ£»½¼0š&šH»¡•Ã¾&ž'»¡½ • ³" " % š  4²• " Ò Ñ ½Ð ž š š  • ¾ ž š  • – ž  &46„›µ46®4t²• “ œ Ó ¦W “ à 4²• š  ³ š  46• ‰¡   " š ˜ y¯• Å š š ˜ • – • ›y™—¡'” ‰¡   Å “ š  046• Ò Ñ Ð ½€hi h h fh Á „¥ €¢ ¤ ¤ ‰¡   “ š ¿ ž Ÿ ž  07Bx4t²• œ ¢ F « I hh Ó ¦W r S š š ˜ • – • ›y¯®¡'” ° p p p p p p qo ØÆÙ Õ Ö Ø6× µÕ Ö Q RP Ô » ¼ š š » • ¾ ž » • )½0›D(Ì&'¡½ ° &y¯®(xǹ’ š š ˜ • – • ” “ ž š  4²• ž š  µ46• ‰Ê ‰Ê ¼ š š  • ¾ ž  • Î Ä  ¼ š š  • ¾ ž  • Ë 0›4²Ã›t²'Ïͨ0&46Ì&t6®¹À46Ã#¼ “ š  • ¾ ¼ š  • – š š  • ¾ ž  • ‘ Ä  ¼ š  • – š š  • ¾ ž  •  ®46®›&46Ã&t6)É£¨›4²—À›46țt²½Ç—4²®Æ¼ “ š  • – ‰¸   š š  • ¾ ž š  • – ž  ›4²@›µ46®4t²• œ ³" " " % š  46• ¢ H§ ¦ © “ š  46• š  4²• Å š š  • ¾ ž š  • – ž  ›46@&4²®4t²• " &y¯®(x” š š ˜ • – • œ « © €D§ ¦ ¢ © ¦ ‰¡   “ ± “ š  ®46• Then ± Take Example 22. .± 1 ± ± 1 1 ¬ ‰C 8 ¬ x ¬ Õ¯ À t± 1 ¬ V G F Î à ± ­ P¬ U ² ± × Ëf¬ ± Ë× ¬ xU ‰C £8 « In every case. Define .

and all other variables are functions of š ¿ ž  „4t²• “ Ä œ  Ó ± where (see (3.° ›46țµ46®4t²• Á š š  • ¾ ž š  • – ž  œ š  • – š š  • ¾ ž  • 46®0›4²Ã&t6)‘ š š  • ¾ ž  • &46Ã&t6'Î Ä š š  • ¾ ž š  • – ž  '&46Ã&4²—Bt6• ¥ œ “ š  • )4²„æ ° 4²• š  ‰Ê ¼ š  46• š š  • ¾ ž  • ‘ š  • æ Å ›4²Ã&t6)›4²„¹“ š  • 4²®– ‚ š š  • ¾ ž š  • – ž  ›46„&46®Bt6• œ “ ‚ š  4²• š  46• “ œ š  46• š  4²• Å  ¼ ž š  4²• ‰Ê å ¼ iÁ œ Î äÄ ¥ œ – 3‘ 㠚  D46• Þ Å Å  ¼ š ¿ ž Ÿ ž  7B'Bt6• ± Ë ÉÄ ¥ œ â ¨“ Á Á œ ß Î ß Ü Ä Á R¥ œ Ÿ Î ‘ D3ƒÄ ¥ R¥ œ ß Ÿ ß ‘ ß Ü Ä Á œ œ œ Ÿ  iÏÄ ¢ œ œ % ‰Ê Á  ¼ ¨0š Á œ ß Î ß Ü Ä Á t¥ œ % – Î ‘ HDÉÄ ¥ R¥ – œ ß ß Ü œ ß ‘ ß Ü Ä Á œ Ë ÉÄ ¼ )š Á œ ¥ œ Î ƒÄ ¥ œ Ä œ – ‘ • 3(¨Ä –  iÏÄ ¢  œ Ü ß Ó • š  4²• Þ Å “ Á ¾ ¼ #á#¼ Á ¾ Á Ä ¨àƒ½¼ t¥ ¾ ¼ – ¾ #¼ Á œ ¥ Ä – ¼ – ½ƒ½¼ R¥ Ä – „½¼ ¥ œ Ä  ¤¨¼ ¢ Ä Ä š4²• Þ Å Ó “ ‚ š  46• š  46• Å  ¼ œ  ¼ š š  • ¾ ž š  • – ž  ¨0&46@&46®Bt6• œ œ œ š š  • ¾ ž  • ›4²Ã&t6½ ­ (¢ ª S š š  • ¾ ž š  • – ž  ›46„&46®Bt6• œ Ò Ñ ½Ð Ý · ›Ü Q RP ¢ º F Ó Ú I “ š  46• š  46• Å Û » ¼ š š » • ¾ ž » • £½0&H¡Ì›x(½ where it should be noted that The partial differential equation satisfied by ‰¸   ± is a martingale under Example 22. . We want to choose š  • 4²„æ š  • – 4²—`– “ š š  • ¾ ž  • Î ›4²Ã&t6'¨„Î “ ± Therefore.3) CHAPTER 22. we should take š š  • ¾ ž  • ‘ &46Ã&t6)`„‘ “ š  • 4²„æ to be . Summary of Arbitrage Pricing Theory . and .2)) . is . We have . We have so that 231 .4 (Continuation of Example 22.

232 .

). (i.. ² ± × ©Ëf¬ ± × Ëf¬ .62 (Levy) Let . this shows that is normal with mean 0 and variance . such that: C ½ C be a process on . We shall show that the ± é Since the moment generating function characterizes the distribution. We compute (this uses the continuity condition (1) of the theorem) ± × Ëf¬ C H8 E so S Q P Ö RtRI « 8 ë Âæ ‹ 5 Ä uses cond. adapted to a filtration Ù ­ ­ ® × ’„dË× ¬ ÃC ² ± è ½ Ù ± é × ¬ Ù C ç ­ ® ± × ’iËf¬ C µ ± ©Ë× ¬ C $ .. We need to show that mean zero and variance . informally is a Brownian motion. Then is a martingale.Chapter 23 Recognizing a Brownian Motion ± € Š® à ± ê 2.e. and is independent of conditional moment generating function of is $ C F é C H8 ¬ C ô ó 1 dÎ ô Ö E $ ê Å à 'Eæ × 8 ± « ¬ ‘ Ä 5 ó ô u² à Eæ ‹ C H8 5 iÄ ± E ô C H8 é ë Âæ % ± ó 1 dÎ ô ± × Ëf¬ ¬ ‹ é C ¬ 5 Ä µ ± ©Ë× ¬ ± $" 1 " ¡E E " " C H8 é à ¬ ê ê ¯æ F 2Î E C C H8 C é ‹ E µ ± × ÌËf¬ Å Ræ à à Ræ E µ Õ× ê ¯æ ‹ 5 iÄ ‹ ‹ æ 5 #iÄ 5 Q² Ä C ± 1 é ½ ¬ ² $ Ù E E ßÞ à Eæ à Eæ × Ü é ‹ ‹ ½ 5 ¨Ä 8 5 Ä µ × Ù é XÙ × ± × f¬ 1. i. with . is independent of . 3 233 ± é ¬ C µ ± × f¬ C E ¬ ® ± × iËf¬ µ ± × ÌËf¬ ± é ¬ Proof: (Idea) Let be given. the paths of C are continuous. $ ®  ¬ ® Ù × Ù ­ ± × ’® Ëf¬ Theorem 0. 3.e. and conditioning on does not affect this. is normal.

and so . we get é Therefore. and It follows that 234 is a martingale (by condition 2).$" " " é " (E E " " é " E ê ²æ ‹ 5 ¨Ä ‹ Ö E ê Å à Í£Ræ ‘ Ä 5 ó ô Q² ² % ± ¬ ¬ $" Å Eæ à ‹ E æ 5 #¨Ä ßÞ ® E 'Ræ ê Å à ‘ 5 óô ç ¯æ ê ‹ E Ä ² ± 5 QdË× ¬ ì % ± à Eæ ßÞ Ö ê ‘ 5 óô Å E ê ²æ ‹ Ä ² 5 „xî © ² ê ‘ Ä î 5 ó ô ec² E ê ²æ ‹ 5 Ä Ö à ‘ Ä î ² ± × 5 ó ô ecŠf¬ ô ê ¯æ ì ‹ 1 óô ® ± × hËf¬ ê ² ± × Šf¬ ì ® |« 8 ¬ ± « à ì F ô 1 óô Î E 5 Ä ¨QŠf¬ ² ± × í ®ì ì E ê ¯æ ‹ 5 Ä ¨Qd± ² é ¬ $" " " é " E ë 6æ ‹ 5 Ä ifŽß QdG£¬ Þ ² ± « ® % °&± ¬ $" " " é " E ê ë 6æ ‹ 5 Ä ¨a!ß Þ F $" " " é " E 1 óô Ö « 8 % &± ¬ à ô Î E ê ¯æ ‹ 5 Ä ¨Q² % &± ¬ à Eæ ‹ 5 Ä ¨fŽß Þ $" " " é " % 4± Ö ± « G£¬ E ë Âæ ‹ 5 iÄ 1 G F !ß Þ é ­ Q² µ q² ¬ Î C D8 à ±$ " « ¬ " " " C D8 E ë Âæ ‹ E 5 iÄ 1 ê 5 Ä 1 G F ê F % 4± ¬ ± « ¬ C H8 ë Âæ ‹ à !ß Þ C H8 ‹ àG  Now ± « ¬ E ë 6æ 5 Ä We define ì 1 so that ì Plugging in .

and are Brownian motions.1 Identifying volatility and correlation CHAPTER 23. Recognizing a Brownian Motion Let and ô Define be independent Brownian motions and and ô ñ The Brownian motions ó ñ ñ ó ñ Then and ô ñ Therefore.Ö ó × 8 ± ô C D8 ô ± tô À ô À ó À ô Î Î ó ó ô À ó tó À ¬ ó tó C D8 × 8ó À ´ó ‚ ò² ô À ô À À ´ ô ² ² ñ ô tô À Î ó C D8 ó ô £ü± ô À ¬ C D8 À ¬ ô ó ñ 8 ó ñ 8 Ö ô ó ® ñ 8 ñ 8 ô À ó À Î Î × 8 × 8 ¾ ¾ ² ² ô ¯ ô ó ¯ 8 ¯ ó ¯ 8 ñ Ö × 8 ² ô ñ 8 ô ñ 8 ó ± ô C D8 ô C H8 ô ô À ô Î ó ó C D8 ó C D8 ó tó ô ó tó C D8 ®ó ’× 8 À¬ ô ó ´ À ¬ ô ´ À ² À ² ² ô ± ô C H8 À Î ó ó ñ ó 8 ñ 8 Ö ô ô C H8 ô tô À ó ó ô À ô À ó Î ó C D8 ô À ó tó À ² ó ² ô ñ 8 ñ 8 Î À ó C D8 C H8 ó Ö ôtô À ô À ô ® tô À óô ® ô ô À Î ó ô À Î Î óÀ ôó À À ô À ï ó tó ² x‚ ² ó ï ð² ó tó ô ô ô À À À ® ô ® ô C H8 ô tô À ó ô À Î Î ó ó C D8 ó ô À ó tó C D8 Î Î À × 8 ¾ ¾ ² ² ô ¯ ôó ¯ 8 ó ¯ ¯ 8 ó C C C H8 × 8 Define processes 23. are martingales. and similarly and have continuous paths. 235 ô ó ñ . The stock prices have the representation and by are correlated. Indeed.

We want to find ‚ ô ñ ó ñ If ‚ This corresponds to A simple (but not unique) solution is (see Chapter 19) so that . we have .2 Reversing the process 236 and are Brownian motions with correlation coefficient . then there is no and ô ñ ô Continuing in the case ´ t ic² ´ t õ² icƒ‚ .® × 8 A × 8 ² ² ² ó ² 8 ô ñ 8 ‚ Î × 8 ô ô µ ‚ Á ñ 8 µ É× 8 9 ô ² ñ ‚ µ ô É£´ A ô ñ 8 ô ñ Î ô 8 ô ‚ ñ 8 ó ´ ‚ µ ô É£´ ñ 8 ‚ Á ó 8 9 ó ´ ñ ô C D8 ô ó C D8 C D8 ® ’× 8 ó Ö ñ 8 C D8 ó ‚ ~² ô C D8 ‚ ~² 8 © C ± ´ t õ² ‚ üicƒ…¬ ô ® C D8 ó ñ ‚ µ ô É£´ 8 ‚ ɵ ô ï C D8 ñ ² Î 8 ó C D8 © ô C D8 ² ñ 8 ñ 8 ô À ÉÌ´ ô ‚ µ ó ó ® ñ 8 ó ô e~² ô À ‚ ó ó C D8 ó ô À À ² ² À Q² Ö ó ô À ‚ µ ô £´ ï ó À ² ô ‹ô À ó ® ® ô e~² À ‚ ó ‹ó À u² À ô À ® ­ eQ² ô ó à ΠÀ Î ó ó ô À tó ó ô ô À À ó ô À ô ôó À ó À ô ³ À À ˜‚ Î Î ó ó tó ô À À À ô ô à tó ô À ô À ó ó tó ô À À ³ ² í ó nó ô Àó À ô tó ó À ˜‚ ³ ¹² à ô tô À ô ô tó ô À ô ô tô À ô ó À ô ³ ¹² ô à ó ‹ô À ô À óô tó À ó ó tó ô À À ³ ² ó ô à tó ô À ô À ® ô ó ® ñ ñ 8 ô À ó 8 Î Î À × 8 × 8 ¾ ¾ ² ² ô ¯ ôó ¯ 8 ó ¯ ¯ 8 Suppose we are given that where 23.

to conclude that ô C ® ± ó ñ 8 ó Ö ñ ­ Qd× 8 ² ± ‚ 8 ɵ ô ‚ Çò× 8 µ ñ 8 ó ó ñ C 8 ¬ ‚ ¯¬ ‚ µ ô Ç£´ ‚ µ ô Ç£´ ´ ´ ô ô ² ² ô C D8 ó C H8 C C are Brownian motions. . cross-variation are independent. Recognizing a Brownian Motion ô 237 so both and ó We can now apply an Extension of Levy’s Theorem that says that Brownian motions with zero are independent Brownians.CHAPTER 23. Furthermore.

238 .

and and are independent Brownian motions on some ô À ‚ µ ô É£´ ± × Ëf¬ ó C H8 ï ó Î C À Î ± × Ëf¬ × 8 ó C D8 ö ÷² C ô e‚ À ± × Ëf¬ ± × Ëf¬ Î d × 8 d ø ~² ± ¬ Ë× °¯ 8 ± × ¬ ËfÕ¯ $ ¬ . we will need the money market and two other assets. In order to hedge it.1 The option payoff depends on both the and processes.Chapter 24 An outside barrier option Barrier process: Ö ¯ ® ± × hf¬ ô d 8 Stock process: C H8 at time . which in this case means the discounted and processes. ô ü 239 ® ô C H8 Î × 8 ô ü ² ô ‰C 8 ® ó C D8 Î × 8 ó ü ² ó ‰C 8 ó ü We want to find and and define ¯ î í eeì Ö ± × f¬ E Ç Ræ d ë ù dBé è ç ± d Ç Ü i­ h û û iG ú ø —}d± ² à ® È Ü ± ­ ¬ ¯ Ü Ô©©P°4x­ È µ Íò± ó . The risk-neutral measure must make the discounted value of every traded asset be a martingale. which we take to be and . where ® § Ü ± ­ ŠÓ¬ d Remark 24. The option pays off: ½ ¬ ¯ °¬ ô ½ ¬ P¥ d ´ ¿b2i%£e„ö Ü ‚ Ü ´ µ ® ­ ¯ d ô e­ À ® ö ±ó € Šµ® ½ À where ®  .

has volatility and zÞ y Under .e. has volatility . ô ó zÞ y ½ ü ± ô ô × 8 ô ô ü ô À ‰C 8 Î Ö ó ô ô ô À ü ó ¬ ô C D8 ô ‚ µ ô £´ ó ‚ µ ÉÌ´ ® ô À µ ò± C D8 ô ½ ¬ ô Ö ‚ ´ µ ï ó ï Ö ó $ À C Î ô ü  Î ü ô À ‚ eɵ ô À × 8 Î ó Î ï × 8 ó þ £d ‰C 8 µ d± ‚ µ ô ÉÌ´ ¾ ó ó ® ü Î ü ó ó ô ˜‚ À ü ô ü ó ô ˜‚ À ô ˜‚ À ½ ü ó ¬ ® z Þ 8 ± ó ó ó ó À À C D8 À C ó Ýø µ À Î ¾ Î ¾ ô ô ˜‚ À Î Î Î Î ü × 8 × 8 × 8 × 8 ó ½ µ ü ¬ ² xö ² Ýø ² q ý ¤if p h ¾ ¾ ¾ ¾ Î ó ô ü ² ² ² ² ÿ ½F ¯ 8 d ¯ d ô ü ² d± ² ± þ dՀ¬ 8 (0. the correlation between and is . ‚ i.2 Under both and . and are independent Brownian motions (Girsanov’s Theorem).2) ½ ¬ ý zÞ ó y ü ‰C ‰C is the unique . ô À ® × 8 ¯ ó ô À À ó À ‚ e~² ¯ 8 ¯ d d 8 ² ± ± ­ òtP¬ d ê ê ˆ zÞ y d ® ± ­ ¬ ¯ ® ­ ¬ hPՒe£§« ù ù ˆ zÞ Remark 24.2) uniquely determine and . risk-neutral measure. This implies the existence and uniqueness of the risk-neutral measure. What matters is that (0.240 so that ‰C 8 E î í Ç Ö ï eeì Ræ We must have (0..1) Ö ô ü ô À ‚ µ ô ÇÌ´ ï We solve to get ® ¾ µ xö ² ü We shall see that the formulas for and do not matter. We define ® s The value of the option at time zero is ë ù dBé è ç Ô·H± ± È µ ½ ¬ ¯ Õ¬ Ç Æ eÅ Ä ä ßÞ y We need to work out a density which permits us to compute the right-hand side.1) and (0.

appearing in Chapter 20. is 241 .ÿ ± ½ ¬ ô ‰C ô À ‚ µ ô É£´ ï Î ½ ü ô À ‚ ˜µ ½ ô ó ô À ô µ ò± ½ ¬ C ô ˜‚ À Î ½ ¾ û p h 3if ± ­ ¬ ¯ P£°Qò± ² ½ ¬ Õ¯ ® h± ½ ¬ C Î ü ó µ qò± ² ½ ¬ ‰C ½ ÿ ± ÿ ô t± À ô ‚ ¬ µ ´ ½ ¬ ô ‰C ô À ‚ µ ô ǁ´ ï ï Î ± Î ½ ô ó ½ ô ¬ ‚ ‰C ô ˜‚ À µ ò± ½ ¬ Î ó ½ ô ‰C ó ô ½ ô µ d± ½ ¬ ô ‰C ô À ‚ µ ô ÉÌ´ ô À ó ô ® ô ‰C 8 ô À ‚ µ ô ÉÌ´ ï Ö Ô ü óô ü £¥ ¨Ü £ » ½  ® Ô £» 8 8 £¥ £ ¤ ½ ô µ £»   Î ô ± » ÿ   £ ½ Á µ d Á ¬ £¥ µ W  p h 3if £¥ 8 ± ½ ¬ ± ± £» £» ¡ ¢ ® ¤8 ½ ¬ ² d± d Ö ÿ ± ® ó ÿ ½ ¬ ¡¢ ó C À gû À gû p h ¤if p h ¤if ± ­ P¬ ± ­ P¬ d ± × Ëf¬ Ö Ç ± × f¬ ® ± × iËf¬ ó ó C h û û iG ú ø —}d± ² à Î ‰C ó   À ·µ × À P” ü ½ ü ¬ C ¡¢   Set ² ± dË× ¬ ¾ ² ® Á ”     Ö ó s R× ô ó À ô ó µ ± × df¬ ‰C ó À Î × ¾ g¤if q p h ± ­ P¬ d ² ± dË× ¬ d ó ® ‰C 8 ó À Î ¾ ² d d 8 × 8 Then   From the above paragraph we have     ó ô À ô µ ô ˜‚ À ¾ ½ Î ½ ¾ û û p h 3if p h 3if ± ­ ¬ ¯ P£°Q² ± ­ ¬ ¯ P£°Qò± ² ½ ¬ Õ¯ Î ó ‰C 8 ô e‚ À Î × 8 ¾ ² ¯ ¯ 8 ® ­ e©ö £¥ ½ ² y ¦ §Á ± µ ñ Á ¬ Á ½ ¬ ± d d £¥ ¼ û C iz Þ ¡¢ ½ ½ C ¬ ¥   ² ± dË× ¬ The joint density of   ¬ so so The stock process. so Recall that the barrier process is and     CHAPTER 24. An outside barrier option .

2)) to determine and . the pair of random variables is independent of are independent under .1) Ö § k ô d À µ k Æ À È t® ¾ ² ü ó ü ô ü ó ® ü ® ‚ C® ô ’® À À   ± ­ P£¬ d ® ø ® dµö ± ­ ¬ ¯ PՒ® ½ Ö ü £¥ ½ º Ö 8 l» y» 8 8 £ ¤ £¥ Ô ½ ô ü óô µ £»   Î ô ± » ½ Á   £ µ £¥ Á ¬ Ô µ W p h 3if ± » W p h 3if £ Á ¼ ½ Ö µ ñ Á ¬ Á ½ ¦ § Á ½ ©µ º ô » F ¦ Á ´ ¼ Ö 4 4 5Å 5Å £» ô ˜‚ À Î ½ ± ü ô ˜Éµ À ‚ ô ó ô À ô µ ¾ ¬ p h 3if Ú ± ­ ¬ ¯ Ó°P F G F   4 Ç Æ Å Ä e#Q² À ® ± ­ ¬ ¯ ® ­ ¬ hÓ°eC« 0 ' $ 2 3 1) (& #"!k k % ± ± ­ tP£¬ d ¡ ¢ ® £» 8  ± ½ ¬ C ® » 8 º  ± ½ ¬ ô ‰C û zÞ y   zÞ y ô ‰C and ± ½ ¬ ô ‰ C iz Þ û y æ ù Bé è Ö Ç Æ ß Þ ËeÅ Q² Ä Ç Æ ˜Å Ä ä ßÞ y ² ± ± ­ ɋP¬ d ® ± ­ ¬ ¯ ® ­ ¬ hPՒeC« . We would have had only one equation (see Eqs (0. any option whose payoff depends on cannot be hedged when we are allowed to trade only in the stock. Therefore.(0. The density of is because ¡ó ¢ ® h± ½ ¬ ® C h± ½ ¬ ô ‰C ¬ ± wÞ  The answer depends on and . The nonuniqueness of the solution alerts us that some options cannot be hedged. the joint density of the random vector is ½ ¬ ô ‰C ½ ¡¢ ¬ ½ C ¬ î í eeì %  E Ç Ræ   E ©  k À G Î ± ½ ¬ C ô e‚ À Î µ ¾ ¬ p h ¤if Ú   ± ­ ¬ P°¯  ¬ ¯ °£¬ ½ y   ó ü ± ô Àeɵ ô À ô ‚ ô ½ ï eeì E Ræ î í Ç   ë ù é ƒB˜è ç Ô·H± ± È µ We know the joint density of 24.1). then we would not have tried to set its mean return equal to . ô ü ó ü ô ü ô À ‚ µ ô ÉÌ´ ï Î ó ü ô À ˜‚ Î Š¾ ² Ýø ¾ zÞ yÖ ÿ º » 8  ± ½ y ¬ ô ‰ C áz Þ û ² ÿ ® Ô 8 £¥  ± ½ ¬ ± t± ½ ¬ ¡¢ ® h± ½ ¬ C ¬ º » º ® » 8 Ô ½ ºÁ ô » µ ‰W p h ¤if ½ ¦ Á ´   ¼ ² ÿ º » 8  ± ± t± ® i± ¬ The option value at time zero is and . The parameter appearing in the answer is Remark 24.1 Computing the option value 242 .d   ‚ ç È Íµ Û º ô À ô É£´ ‚ µ » ï Î ÿ £¥ 8  ± ½ ¬ ¡ ¢ ® £» 8  ± ½ ¬ C û   ± t± ½ ¬ ‰C   ½ ¬ ô ‰C Ö ‚ ç È Íµ Û ± ½ ¬ ô ‰C ô À ‚ µ ô ÉÌ´ ï Furthermore. It does not (1. It also depends on depend on nor .3 If we had not regarded as a traded asset. Indeed.

ü and we are on our way. we can show that the stochastic E A C A@ FDB 8 6 97é ® è ç ± È µ ƒtÔÍò± yñ ë B˜ƒtÔÍò± ù é è ç ± È µ ô À Î ½ × 8 ¬ ¯ Õ¬ ä ¾ ² u #ã à â â ¯ 8 ½ ¯ ¬ ¯ °£¬ ß Þ eIE˜¥QòGR® à Å Çæ Æ Å Ä ² ± e y E ½ ® ñ « Ø Î ® × ¬ fC« ¾ ã « ©¯ ô ˜‚ À µ 9 He#Q² à Æ Å Ä ü ½ ² Î yñ so with we have Ö ñ ô À ® ¾ ô À Î µ bø × 8 ø ~² ² ü ¯ 8 ¯ 8 × 8 8 ñ so is a Brownian motion under (Levy’s theorem). Using the Markov property.CHAPTER 24. We compute ± ± × tËf¬ d ® ± × ¬ ¯ ® ¬ à Æ Å iËfՒ’× §« e#Ä ± ­ P¬ d and ± ± ­ tÓ¬ d ® ± ­ ¬ ¯ ® ­ ¬ iP°eC« e Ø × Ôµ ± ­ P¬ ± ­ ¬ Ó°¯ d by replacing . and Now start at time 0 at process ± ­ ¬ P£°¯ by . ï is superfluous. namely. and ® ô C D8 ô À ® ô ‚ µ ô É£´ ‚ µ C D8 ô ÇÌ´ Î ï ó Î C H8 ó C D8 ô ˜‚ À ‚ ~² Î × 8 ñ ø ~² 8 ¯ 8 ¯ ¯ . and respectively in the formula for . there will be only one . 24. An outside barrier option d 243 . Returning to the we should set zÞ Now we have only Brownian motion. then ¯ ã hã Ç Ræ 8 8 d If we have an option whose payoff depends only on original equation for .2 The PDE for the outside barrier option Returning to the case of the option with payoff ® î í eeì E we obtain a formula for ï ® î í e%ì E 5 6æ ù × 8 A u u « ô d ó ô ó À ô Î u iã « d ¯ ô À % ó ó À e‚ ‰C 8 u Î « ó « ô À ó ¯ ô À ô ô Î ô ‰C 8 Î ã u « ©¯ ô À « ¾ d 4Î ‚ µ ô É£´ ã « Š¯ ¾ jÎ à ï ± ± × tËf¬ ï Î « ó d ® ± × ¬ ¯ ® ¬ à Æ iËfՒ’× §« eÅ Ä ä ‰C 8 Î zÞ y is a martingale under 8 . .

L) = 0. 0) = 0 ® "­  Ç  £½ Setting the term equal to 0.1: Boundary conditions for barrier option.244 y L v(t. Note that is fixed. we obtain the PDE The terminal condition is and the boundary conditions are Ö § ©Ù e Ù ­ ® ­ e4ý Ø × ® ® Figure 24. § ½ Ö Ü ~e Ü ­ ©ý × ® ­ eQ² Ù Ù ® Ø ­ ­ ® ­ de©ý ½ u u Ù ® « × ô ½ e Ù ó ô Ù Ø ó dÎ À ô ­ × ® Ù ® ­ "QIPe’f§« ² ± ­ ® ­ ® × ¬ ç Ô·µ ± È ­ u iã « Ge ® ­ eQH± ² Ø ô À ã hã Ø ¬ ñCGR® ² ± e « ó ô § Î Àe‚ ó Ø ô À ô ô ® Ø ® × ¬ f§« Ø ® Î u ½ ¬ C« « e ¾ ©Î ã « ¾ Ø jÎ à × 8 « Î « ¾ µ . 0. x >= 0 x v(t. x.

An outside barrier option This is the usual Black-Scholes formula in .3 The hedge term to 0. The value of the portfolio has the differential ± × f¬ × ± Ë× ¬ ô V Ö d  à Æ ˜Å Ä  8 u « Î  à Æ ¯ eÅ Ä  8 ã « u² ï ± ± × tËf¬ d ® ± × ¬ ¯ ® ¬ à Æ iËfՒ’× §« eÅ Ä ä 8 Ö ± Ë× ¬  ï d 8 ó ± × f¬ Î ó ‰C 8 ± × f¬ d ¾ d ó à Æ À eÅ u² Ä µ  × 8 f¬ ± × à Æ eÅ u² Ä à Æ Å Ä e#u² ï ± × Ëf¬ d à Æ eÅ Ä ä 8 Ö ± × % f¬ ô ‰ C 8 ± × ¬ f°¯ ô À Î ‚ µ ô ÉÌ´ ± × Ëf¬ ‰C 8 ± × ¬ f°¯ ô ˜‚ À × 8 ËfÕ¯ ± × ¬ ¾ µ   ± × ¬ ËfÕ¯ 8 Î à Æ eÅ u² Ä ï à Æ ± × ¬ ËfÕ¯ ËeÅ Ä ä 8 ® ¯ d® ô ‰ C ® d ‰C ó ® % u ‰C 8 ó À Î « ô ‰C 8 ã « Š¯ ô À ‚ µ ô Ç£´ ï Î ï ó ‰C 8 ã d « Š¯ ô e‚ À à Æ eÅ Q² Ä ± ± × ‹Ëf¬ ® ± × ¬ ¯ ® × ¬ à Æ hf°’fC« eÅ Eä Ä 8 × 8 Ö § ©Ù e Ù ­ Ö ­ ©ý Ø ® ç Ô͵ ± È E Ø ¬ d§Ó’® ² ± ­ Ø ® þ ­ eQ² ç ÔÍ%¬ ± È µ ­ à Å Çæ Æ eHReÅ QHP’e’fC« Ä ² ± ­ ® ­ ® × ¬ þ ­ eQCPe’fCŠeQI± ² ± ­ ® ­ ® × ¬ « ® ­ ² ­ Q² ã hã « ô ó Ø ô À ô ô Î ã « ¾ Ø jÎ à « Î « ¾ µ ­ uše ² CHAPTER 24. ² ã « ¬ C« e ¾ µ ² Ø 245 . the barrier is irrelevant. the option value On the boundary. and let denote the number of “shares” of the barrier process . we have the equation . This is the usual Black-Scholes formula in . and are functions of . Note × d ó ± ± × tf¬ d ® ± ¬ ¯ ® × hË× °’f¬ u « ² u « ± ± × tf¬ d ® ± × ¬ ¯ ® × hf°’f¬ ã « Therefore. and the option value is given by the usual Black-Scholes formula for a European call. ­ ² be ® ­ eQIt"’’× §« ² ± e ® ­ ® ¬ Ø ­ q² ® ­ ® × ¬ e’fC« Î « ­ where that After setting the 24.Ö × 8  ó V d ó V µ Ô¯ ô V µ U ¾ ´ ©Î d ó 8 U V Î ¯ 8 ô V ² d U 8 ± Ë× ¬ Let denote the number of shares of stock held at time . Ø ­ Q² u u « ô e the terminal condition is The boundary conditions are the terminal condition is ½ ¬ C« Ö ­ ý ©xe ® ­ eQ² ç Ô·ÕñCte® ± È µ ­ ¬ ² ± e ® ­ ½ § ó ô ó À ô Î u « e ¾4Î à « On the is boundary. The boundary condition is .

Ö ± ± × ‹Ëf¬ d d ® ± × ¬ ¯ ® × hf°’f¬ ® ± × ¬ ¯ ® × hf°’f¬ u ó « QdËf¬ ² ± × ã « Qdf¬ ô ² ± × V ® ± ± × htf¬ V ± ± ­ tP¬ d ® ± ­ ¬ ¯ ® ­ ¬ « hPՒe£§QòP¬ ² ± ­ × U d U Ö  ± × Ëf¬ d à Æ Å Ëe¥Ä  ó ± × 8 Ëf¬ V Î  ± × ¬ ¯ à Æ Å f°e#Ä  ô ± 8 Ë× ¬ V ²  ± × Ëf¬ U à Æ Å ˜¥Ä  8 To get ± ± tË× ¬ ® ± × ¬ ¯ ® × ¬ « hËfՒf§udËf¬ ² ± × and This is equivalent to 246 for all . we must have .

2 First passage times for Brownian motion: first method (Based on the reflection principle) ­ ©ö Ø zÞ C Let be a Brownian motion under . let be given. Suppose we exercise the first time the stock price is Ö ç tf°Íµ ± ± × ¬ ¯ È ¬ | Intrinsic value at time  ¯ IÀ Î × 8 ¯ ¾ ² ¯ 8 G H× or lower. 25. We define .Chapter 25 American Options This and the following chapters form part of the course Stochastic Differential Equations for Finance II. § ® ç ± tt± ÿ § ©Ù í IÆ % PeÅ ÷ ¬ ¯ °Íµ ± × ¬ ¯ þ ­ ËfՒe4ý þ ­ C e4ý Ä ß t± Þ È ¬ § × û ú ù ø üåiQi÷ ² Ø IÆ Å Ä Þ ² % 7˜¥iß Qd± µ V¬ È × û ú ù ø üåiQ² µ È W ² í Ø ÷ ¬ í « È ® ’"­  Ç`§  Let be given. We compute the distribution of . We need Ö § Ù ©bØ § ö Ø if if . and define Ö Ø ÿ ² ± × dËf¬ is called the first passage time to .1 Preview of perpetual American put C D8 25. 247 ÷ Ø ÷ § The plan is to comute and then maximize over to know the distribution of . í ± ÷ Ø ¬ í « to find the optimal exercise price.

From the first section of Chapter 20 we have 248 Figure 25.® Ø ý ± × £f¬ ¢ © „§ × Ù ÷ Ö S 8 Ô Á ô S µ W p h 3if ¦ Á ¼ Á à T à U Vã 4 3T F ² Ö " " Ô ¥ 8 Ô × Á ô ¥ µ µ W p h 3if × ¦ §Á ¼ Á × ã 4 ã ¼ F F ² ¥ 8 4 5Å w R "" 3 wR × Á × Á ô ± » µ W µ ñ Á ¬ p h ¤if ¥ W p h 3if ¦ §Á Á 4 F ² ¥ 8 » £8 Ô × ± » ¦ Á ¥ ¼ × 4 5Å 3 ã F ô ± » µ ¥ Á ¬ µ d Á ¬ Á 4 ² Ö ¥ QÜ ® ­ » e©ö ¥ ® 8 » ¥ 8 Ô × Á ô ± » µ ¥ Á ¬ µ W p h 3if × ± » ¦ Á ¥ ¼ × ² µ d Á ¬ Á Ö ± 1 à ¬ C hû 5 û iG ú ø @ʊË× ¬ ² ± ¢ Define We make the change of variable ² S ÿ CØ ý ± × ÌËf¬ ¢ û áz Þ ÿ » 8  ± × f¬ ® C Ô 8 ¥  ± × Ëf¬ ¢ û áz Þ Now Therefore.1: Intrinsic value of perpetual American put K in the integral to get K Intrinsic value Stock price x .

Brownian Motion and Stochastic Calculus. Brownian motion and Stochastic Calculus. pp 196–197.Ö Ù ½ × Ü ji­ ® ’× 8 Ô × Á ô Ø µ W ÿ p h 3if × ¦ Á Ø  ¼ × ² × 8 ÷ û Þ gáz u² Ù ½ ² ± þ d°™¬ × zÞ y ÿ × 8  º ÷ iz Þ û y ­ ® ± × ’hf¬ ‰ C Ö ± ½ ¬  þ ® z 8 ± ½ ¬ ý ÿ F ½ ½ Ö Ø ÿ ² ± × Šf¬ ‰ C "©ý þ ­ º × CåiQx÷ û ú ù ø ² ® ® ÿ ÿ × × ô ô ü óô ü óô Î ± × Ëf¬ ‰C C ü ü µ Gû µ Gû Î C p h 3if ² ² ± × ©Ëf¬ ² ± × df¬ ‰ ý C µ ± ©Ë× ¬ p h 3if ® ± hË× ¬ × ü Ö ÿ ­ ©ö  Y ® X ô T ã eÅ Q² Ä × 8 à ÷ û gáz Þ ¥ Å Ä X ` 4 G F ² I X Y5Å Ä iß Þ Ö × 8 × Ô × Á ô Ø µ W p h 3if × ¦ §Á Ø ¼ × ² × 8 à ‚ × Ø Ô ¼  W Ö Ô × Á ô Ø W Á µ W p h 3if ¦ Á ¼ Á F v× µ q² × 8 S 8 ô S µ W p h 3if ¦ Á ¼ Á à T U Vã 4 ¢ W û —z Þ ÷ û gáz Þ × × W W W ³ ¹² × 8 ÿ CØ ý ± × f¬ × 8 ÿ × Ù W ² W ² ÿ × 8  ÷ û gáz Þ so For ª Ü × Reference: Karatzas/Shreve. More specifically. the process . American Options . define Define We also have the Laplace transform formula CHAPTER 25. pp 95-96. so $ Ù zÞ y (See Homework) ”. Under ½ We fix a finite time fixed. define and change the probability measure “only up to .3 Drift adjustment Reference: Karatzas and Shreve. is a (nondrifted) Brownian motion. 25. with 249 Ù ­ .

250 is arbitrary. Recall the Laplace transform formula for .4 Drift-adjusted Laplace transform we have . this must in fact be the correct formula for all ­ 4ö × For for nondrifted Brownian motion: 25.® ÿ Ø ² ± × Šf¬ C Î × ü þ ­ ý × û e©üú ù º ø Qx÷ ² Ö ­ 4ö Ø ® ­ e4ö Y ® C ô X T ã Å Ä ˜¥QÉ× 8 ² Ô × Á ô Ø µ Š× Y µ W p h ¤if × ¦ Á Ø ¼ × 4 G F ² ‘ I X Å Y5¥Ä ßÞ ÿ CØ ² ± × df¬ þ ­ ý × û ú ù ø ² C e©ü&åiQx÷ ½ Ö Ù ½ × Ü ji­ ® ’× 8 Ô ô ± Ë× ü × Á µ Ø ¬ µ W p h 3if × ¦ Á Ø ¼ × ² ÿ × 8  º ÷ áz Þ û Ö é 8 é Ô 8 Ô é ô Á Ø ÿ #é 8 ± é é ½ü Á µ ü óô Ø ¬ µ W p h ¤if é é ¦ Á Ø ¼ é é G F à ¼ Ø G F ² é µ ô ô  µ Ø #ü W p h ¤if ¦ Á Ø #ü û à p h 3if G F ² º ÷ áz Þ û ÷ ô ü óô y #é ÿ ô Ø gü ü óô µ à ßÞ ßÞ ßÞ ßÞ ßÞ ß u² Þ y y y y y ² ² ² ² ² ² ÿ × Ù º ÷ áz Þ û Ù ½ ï ÿ º ï ÿ º % ± × ½% b¢ µ º ‰C û ü p h ¤if î à p h ¤if î à º ± Ë× $ " Ì÷ ¬ ¢ " " ô ü óô ü óô µ ± dË× µ d± ½ ü óô ¢ Ì÷ ¬ û ÷ ¬ " ÿ ½ ô ï ÿ ¬ µ d± ‰C ü û ¬ ‰C p h 3if ü ßÞ y î à ½ ô ½ û ± ½ p h ¤if î à û I û beè ä a é I û bé è ä a I û ab|è é I a bé è ä û û % ¬ ´ ý î à ï î à I abé è I a é beè ä For × Ü ji­ Since Therefore.

Then . a I `¥Ä X Å ø åù Ð (Recall that ² d± Letting ­ e Y  º ¬ I÷ If ­ ©ý ü If we obtain 25. American Options with replaced by Î Y Y ßÞ Y ® ‘ c ç dPX ô Ø × ¼ × T 4 ã Å e‹ ã Q² Ä G F c V p h 3if × ¦ Á c Vã G F Ä Q² µ W p h 3if ¦ §Á Ø × ¼ × 4 ¼ × ² µ £W p h 3if ¦ §Á Ø 4 G F ² a I 5Å X Ä iß Þ If . then . so . . then and using the Monotone Convergence Theorem in the Laplace transform formula ). then for every ­ ©ö ­ u² # æ ª Therefore. 251 be given.5 First passage times: Second method ­ 4ö Ø ­ ©Ü Let ü ­ ö ©ÔÀ (Based on martingales) . then a I 5#Ä X Å G e YX ø ù Ð a I `Å X Ä ª Ü £±  º ¬ I÷ if .ÿ × ô ó À ô µ ± ©Ë× ¬ À C gû p h 3if ² ± × df¬ d Ö ´ QÜ c Vã ô ´ q² Ä Q² ÿ 4ª º Ü õ÷ áz Þ û Ö ÿ 4ª º Ü b÷ áz Þ û Ö f f ã ÅcãÄ Dc etViQ² ‘ c T T ô ® ‘ ã Å ã Ä e‹ hQ² c V ÿ 4ª º Ü õ÷ áz Þ û c ç !ÓX ã Å ã Ä et iQ% ² c V a I 5#iß Þ X Å Ä e YX G Ö 4 ì I a è ² E # æ Ö ­ Q² E # æ a I 5Å X Ä G e YX ø ù Y Ð E þ ´ Pq² ü óô E # æ ô I X Y5Å Ä Ø Y ® ­ e©ö Ô ® ­ "©ö × 8 Ô × Á ô × ô Ø µ × É± ô ü óô Î ¬ šµ W × 8 ü óô µ ü 'Ø Îš× Á ô Ø µ d× Y Y × 8 Ô ô ± × ü × Á µ Ø ¬ µ d× the Laplace transform is where in the last step we have used the formula for CHAPTER 25.

ggg p Ö r c ± × f¬ C p ih Î × S ‚ Q RP Á À Î µ gggi h ¾ À  I fhh ggg À ‰ p g †gg ˆ p h f 3inØ ² p qo ÿ ± Ë× ¬ C À × ± ô ó À ô µ ¾ ¬ ˜û Î p h f 3inØ ² ± × ¬ df°¯ Ø ¾ ² ± ­ ¬ òPÕ¯ ² ¯ 8 C D8 ¯ CÀ × 8 ¯ ® ­ e4ö Ø ® ­ e4ö ÿ ± ÷ ¢ × eb0f¬ ô ó À ô Ö ­ ² ÿ × ô ó À ô µ Ø À gû p h 3if 4 Ö I ‘ À ó ô Å #ã À Ä Q² ÿ ª ó À ô ± ÷ ¢ eÝ£× ¬ ô Ü ±  Ö ÿ ± ÷ ¢ × eõ£f¬ ô ® ã ehÄ Ù ¿ÿ ± ÷ ¢ eb£× ¬ ô Ö ÿ ± ÷ ¢ × eõ£f¬ Ö ÿ ô ó À ô ± ÷ ¢ × eÝ)f¬ for the first passage time for nondrifted Brownian motion. . For those for which to be zero if ª ô ó Ä Let 25. .6 Perpetual American put T ã Å e¥Ä . We want to take the limit inside the expectation. We have for which ¬ I÷ ø åù Ð ¨ à ß QÕ´ Þ ² p h 3if Ù ­ 4 ø ù Ð ¨ à ² ß Q² Þ p h ¤if d ß Q² Þ d ² Õ´ ± ÷ ¢ × eb0f¬ d where we understand 4 ø åù Ð ¨ à  Therefore. Y ®  I X Å Ä Þ ² Y5#iß QiX ô ² ÷ ® Á ¼ À Å #ã I ‘ À ô I ‘ À ô ó Å #ã À Ä À ø åù Ð Å #ã Ä Þ iß Q² % 4 ì I è I ‘ À ô p h 3if ó ß Q² Þ ¨ à ß Q£´ Þ ² µ ± ÷ ¢ × deb0f¬ À C #û 4 ø åù Ð ¨ à Ù šÿ ± ÷ ¢ × eb£f¬ ô ó À ô µ ± ÷ ¢ × deb0f¬ ª p h ¤if À C gû ² d±  p h ¤if ó À ô µ ± ÷ ¢ × deÝ0f¬  À C #û 4 ø åù Ð ¬ I÷ ¨ à µ ± ÷ ƒeb¢ × ¬ À C gû p h 3if 4 ó À ô µ ± ÷ ¢ × deb0f¬ À C #û p h 3if C ßÞ À gû µƒ±eb¢ × ¬ C gû ÷ À ó À ô µ ± ÷ ¢ × deÝ)f¬ ô ± ÷ ¢ × eb)f¬ ± ÷ ¢ ­ ebg¬ There are two possibilities. Therefore. We have again derived the Laplace transform formula . so Y ² bÀ ó À ô ² Y . Since is a martingale. so 252 is also a martingale. . For those this is justified by the Bounded Convergence Theorem.

® ‘ ÍÀ U Æ ô § ± § µ È ¬ ñ² y w xí ã Ö § ® ý ­ § Ù Ø ©õ©Ù ‚ ” Á £À ‚ ô ” Á )À u F” u Y” ô À ‚ ô À Î ¾ 4Î Î ¾ µ ô À ¾ ô À ¾  ô ” À Á )·µ Û ô ü Î ¾ Á ô ÿ Ø Intrinsic value of the put at time : We want to choose the largest possible constant.   ÉT§ We compute the exponent Define CHAPTER 25. 253 . . The constant is ‘ 'À U €Æ ô ‘ À U Æ ô Ø ¬ Å v 9± § Ø ® ‘ À U Æ ô Å w xí ã v 9± § Ø ® h± µ V¬ È µ V¬ ‰ ˆ È † ² d± í « Î À ¾  ´ À µ ó ô ó µ ô Î Ö ô À ¾ µ Ú² Á ô À ¾ ñÚ² µ Î ô À ¾ Î ÎÔÀ ¾ Š s tÀ ´ µ Ú² ô À ¾ ô Î ¾ Á Î ¾ Á s À ´ µ À ó µ ó µ ó ô ô À ¾ ñÚ² µ Î s ´ ô ô À ¾ Î µ Ú² Î Š¾ ü s tÀ ´ ô ô À ¾ µ Ú² ô À Á ô ´ µ ü À µ Ö § Ø ‘ ‚ § Ø  p h 3if c ç dtÆ ô µ Ø Ò Ñ PEÐ ´ À § T kÀ Å À ü qÀ rÅ µ Ú ± § ± § § µ V£ñ² È ¬ Ò Ñ PEÐ µ V£ñ² È ¬ µ V£ñ² È ¬ í « IÆ % 7eÅ Ä Þ iß t± § Ø ÿ § Ò Ñ PEÐ ´ Ò Ñ PEÐ À À ´ ² ± × Šf¬ C µ ð× Î C ü µ þ ­ £"©ý × ü × &ú ù û × &ú ù û × &ú ù û ø Q² ø Q² ø Q² í ÷ ² ± × ©Ëf¬ ÿ § þ ­ "©ý ² ± × ¬ ¯ þ ­ df°Š"©ý § ý Ø  ç tf°—V¬ ± ± × ¬ ¯ µ È × Let È ® ’"­ The curves Therefore. American Options be given. Define for are all of the form ® Å Ø y µ È £¬ . .

L) (x/L)-2r/ σ K Stock price x Figure 25. .3: Curves.2: Value of perpetual American put value C3 x C2 x C1 x -2r/ σ 2 -2r/ σ 2 -2r/ σ 2 Stock price x Figure 25.254 value K-x K 2 (K .

American Options we have 255 .Ö ± Ø ¬ « ë í ø åù „ ƒã Ð ² ´ µ %Ú² ¾ ¾ Á ¾ Á ô È Î ¾ Î Á ô Á ‚ ¾  À ¾ ¾ À ¥ Î Á µ ¾ Î ¾ È Á § ´ ± ¥ Á ô À ô À Î ô µ § µ À ¸ í Á Á ô À ¾ Á µ Ú² È  ô À ¾ Á µ Ú² Ø ¬ í È ¬ ô À ¾ Á µ Úd± ² « ë í ø åù e ƒã Ð Ö ¥ § ö Ø ® ­ ó Å ® ¥ § Ü Ø ©Ù ‘ xÀ U Æ ô Å Ø ‘ 'À U €Æ ô ± ¥ § ¬ ¥ ± § µ È £¬ Æ ‘ À ô µ W ® ´ Ó%µ ¾ Ö Á È ¾ Á Î ² d± Ø ¬ « í ô À ² £¥ § ® ¥ ® ¥ § ý Ø ® ­ § Ù Ø ©b©Ù ‘ 'À U €Æ ô Å w ë ã í v 9± ¥ § µ Ø µ È ¬ † ® h± È ¬ ‰ ˆ ² d± Ø ¬ C« Ö È ©Ü § Ü i­ ® ¾ Á ô Á ¾ Ö Á È Î ¾ Á ô À ² § § ­ Q² È ô À ¾ Á Î ‚ ô À ¾ Á Î ´  µ Ö % § È § ´ % ô À ¾ Á ± § § ± Î ‚ ô À ¾ Î ´  ecµ Î ´ %µ Î ‘ À ‘ Á µ V¬ È § ô À ¾ Á ¾ ‘ À ‘  ‚  ‚ § ² § ² § ‘ À ‘ ó Å ‘ À  ‚‘ µ V¬ È ô À Á § µ q²  ‚ y W W and Since Î À ©Ü ¾ We have Note that Ü x­ where to get We solve Solution to the perpetual American put pricing problem (see Fig.4): CHAPTER 25. 25.

5).4) In other words. solves the linear complementarity problem: (See Fig. then K Figure 25. n k ­ ®’ ~x¬~›p d q … q ´ ‹³’ ‘ ² t’  ˆ ­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  … ~x›k k… § ¨† x ~ d© k ¢ ~ «q   q ’ ©o k ’ ‘ ¢¦™ … p¤1o ¡ ¤k ™ ¤1d¨† k n p kª§ n    q ’ p51o1#¡n ’ k ’ ‘ ¢£™ p5k11¡¤k ™ 51dn k ¡ o n p ko    q .f ˜  gY q † (7. If . then . then p d xk ™ 5k o p o q  q r˜  q  † … 256 25. L* K-x K Stock price -2r/ σ (K .1) value . 25.7 Value of the perpetual American put Set Ÿ b– • p o (h ž ” f † œ  œ š ˜– • p h HV›™—ke† §1o “ † “ ’  Ž ” D‘`Yˆ ŒŠ‹p – • ‡ ‰ o ˆ7(~ qr˜ … ‚ € †„jƒ 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’“‘  ˆ † §‰‡… f g˜ … †  ™ ’ §ˆ 9“‘ ˜  Yˆ † .4: Solution to perpetual American put.3) (7. then If – • mQjh l k i If – • mQjh l k i If t l k ¥¤Qi – • where .2) (7.L* )(x/L* ) x 2 (7.

the owner of the put should not exercise (should “continue”). f p o œ Ê ¬Æ p o œ “ or equivalently. If the stock price is in . and is the boundary between them. American Options 257 Figure 25. the owner of the put should exercise (should “stop”).8 Hedging the put œ œ § p h 1o “ – • Á ¼ ” Ÿ hQ¿ ¡¡ n k ’ ¡ ¤k ž ’ ’ ‘ ¢ q n Âq n  ¾k Á ” Ÿ ‰ p5k oÀwp51odn—¾k š½† ¿ k ž ¼ qr˜ š ½† ™ p o œ “ Æ p osÄ p QÅ o Æ † œ œ p o p ut ” h The half-line is divided into two regions: ” ‰ p¤k o n qv ” ˜ h  ¡¡ n k ’ ¡ 5k ’ ’ ‘ ¢ q n »q n    kµ – • ºgk † n ˜ €¹ ·· ¸ sk · © · – • · ·· – • ·· ·· ¶ ˜ (a) (b) (c) . holding shares of of this portfolio is governed by œ ɇ o ɇ p o ɇ p ÈÅ 7(~ o ‘ p sÄ P(~ ™ Æ p o 7(~ q Æ œ … œ œ § … † œ … “ 25. If the stock price is in or at . . Invest the money. One of the inequalities (a) or (b) is an equality. The value .CHAPTER 25. Sell the put at time zero for stock and consuming at rate at time .5: Linear complementarity For all . p sÄ o ” œ Æ p o œ § p p o p osÄ p ÈÇo o Å q Æ q  œ œ œ œ “ p QÅ o pp h o œ Ão dn Let be given.

As long as the owner does not exercise. to be the initial stock price. is the smallest process with properties 1 and 2.2) for . then œ Æ (8.1 If . 3. you can consume the interest from the money market position.e.3) Ó Ò ŽÐ P5(ÃÑÉ p o p p o 1p o É ‡ o¡ n Ï Î ŠÉ‡ Œ 77(~ ‘ P(~ ™ Æ Ê ¬Æ f ˜  Y q œ œ œ Ʌ ‡ p … o œ p o o ¡ p o ‘ 7(~ ™ † p n Ê Æ “ “ ’ œn œ œo ’ É ‡… p p o o ¡¡ p o n Ìp 7p o dn 7~ o ɇ ¢ p “p ’ ’ ‘ £™ p o o ¡ p o  ™ Íp o !n “  q 7(~ Ë ±Æ  … p p o Ç¡§n p sÄ o o Ðf œ ”  Ó Ò(ŽÃÑÉ ÎŒ Š † pœo Ï † œ § „ “ Y ˜ œ œ “ œ f “ … œ “ œ “ œ “ œ “ œ “ The discounted value of the put satisfies … e† œ “ “ œ . we can take to be the initial time and adapt the argument below to prove property (8.1): p o If is not zero.258 We should set To hedge the put when .. f d q ˜ † p p o 1¡n o p mÄ o † œ œ “ f  Ó ÃÑÉ Ò ŽÐ Ï ÎŒ ¬Š ˜ Y ”p o œ p o † œ § “ – • l‘p o œ p Íp o !n “ o qr˜ l “ – • Hp o † œ Remark 25. short one share of stock and hold in the money market. .1) to prove (8.2) (8. i.. p Íp o !n 7(~ o ɇ ɇ ‰ Íp o p o P(~ p Íp œ o “ !n 7(~ … o ɇ qv˜ œ … p Q Íp œ o o É ‡ … “ “ !n 7(~ œ … p p o dn 7(~ “ o ɇ Properties of : is a supermartingale (see its differential above).e. f pp h o 1o !n “ † pp h ‰ 1o “ Ð Ù¢ × ƒØÏ z p ho o |{Dy 1¤Ô qژ  – • w1o i p h Case I: We have Ô Proof of (8. . assuming is a supermartingale satisfying (8. i. Let t Ql i jh 2.2).3). be a supermartingale satisfying We use (8.1) (8. œ 1. and then f t ¥l œ œ p“ h o Íp1o !n 1sÔ p ho f  h “ † ֜ ijh ” p o o!n‚É7(~ p ‡ p ¤Ô o … Q œ œ “ Then property 3 says that f t Ql œ i Õh ” ‰ p o p œ “ ‡ o É7(~ p sÔ o qr˜ … Q œ Ô Explanation of property 3.

t Ql l jh ” p o ñé P(~ p o ɇ o ɇ “ !n 7(~ 2. is the smallest process with properties 1 and 2. ’ ‘5Yˆ   Ž … ï‚ € (Dë† í ìê î Šp Ü o Œ D¯Û ‘” ¤Ô ˆ ” p ko 51dn „ à €  ¦ Q „ Ý € ß 1sÔ p ho •– †ØÞ n  œ “ œ “ t Ý ¥¯Û Now let to get h Ô (Since  f ’ ‘5Yˆ   Ž o Œ ŠDp¯Û ܑ” ‡Ô „¦ € p h o ptÛ Ü‘” o‡Ô  Q 1sÔ  Q – € • —§1o ¿ p h h ¿ uÀÛ Case II: “ : For . 25.10 Perpetual American call ˜ »q p ” o 57~ o ˆ‡ “ … î ‚ € (ë† í ìê Theorem 10. ” ð€p o xé 7~ p ” o ˆ‡ “ Value of the American contingent claim: pp h o Í1o !n f Ó Ò † “ æ çå͐‘5ŽYˆ è ˆ‡ Œ Š ‰ p p | {z y ” o r˜ o 7(~ q … áãä € P⁠Q ’ ‘5Yˆ “  Ž Š o Œ Dp ” ‡Ô (Fatou’s Lemma) (by 8. 7.CHAPTER 25. œ œ “ … Q pÍp o œ pÍp o œ p Íp o o ɇ “ !n 7(~ 1. Optimal exercise rule: Any stopping time Characterization of : which attains the supremum. is a supermartingale. . we have (Stopped supermartingale is a supermartingale) ) … … … .1) (See eq. American Options 259 25.63 ð ‰p f h  k !ò k † p ko 51dn ˆ p ko 51dn o ɇ !n 7(~ 3.2) .9 Perpetual American contingent claim p Íp o xé o Intinsic value: where the supremum is over all stopping times.

260 1. we choose. f p ko ¤1dn † k¨iÂ7p ” o ð “ ˆ‡ 7~ … î ‚ € l Qð ‰ p ˜ «q p ” o 57~ o ˆ‡ ” “ … î ‚ € f ‡ p o É7(~ p ¤Ô o … e† œ œ and define k pÃo h k † p¤k1!n “ o  ’ ɇ 7~ k f q ’ g˜ † É 7‡(~ p o 7(~ … ɇ ˜ q ’ € ɇ … p … p o œ o “ 7(~ †„ ‚  e† ˜«q € p ko ɇ p œ o “ o 7(~ … „ ‚  Q 51dn ‰ p ˜ «q œ “ …„ ‚ € †jƒ Q œ Now start with “ œ Let t ¥Ý Then: Proof: For every . while ) n n”É n” n “ ô Yˆ † 5k ” dn p o † œ Let p h §1o be given. t Ql i jh Therefore. to get . Remark 25. .2 No matter what Intrinsic value: Expiration time: 25. Then has a . œ Ô pÍ1o !n §1‡Ô p h o p ho  ” ‰p ‡ p “ o o É7~ p ‡Ô o ˜ «q … Q œ Ô œ “ 2. It can be shown that jump. k Value of the put: ÐÉ ~ ‡ ‰ p p ” o r˜ o Ñ(Yˆ Ï ~ f q … ‚ € k p “o † œ œ “ See Fig. 25. (value of the put at time if stopping time is a martingale). are continuous across the boundary.6.. is a supermartingale (in fact. i.‚ ƒ‚ ‚ ó Dˆ z ó | Y(1É y { ß í ì (ê ‰ p o p o q r˜ œ ¥eÛ h“ ¿ .11 Put with expiration . f p ko ¤1!n  . There is no optimal exercise time.e.

Figure 25. . q ژ Û “ ” ui p p o ” !n 7(~ o ɇ œ œ … o É7‡(~ p p o “ ” !n 7(~ o ɇ … Q œ œ … i Õh ” p o “ ” !n 7(~ p o ɇ 2.12 American contingent claim with expiration Expiration time: 3. 261 (b) (c) (a) – • . American Options 1. At every point . .© ” ö !­ Û h © ¸ p —¤“ k ” o œ ” p¤1oxé k n É ” h  n k ’ n ¤k n n  q  ‚ ’ ’ ‘ ¢ q ‚ «q  ƒ‚ f p p ” o x1õYˆ (~ o éÐÉ ~ ‡ Ï “ ß … ‚ € ó ˆ ó Y(1É í ì (ê † p o 5k ” dn œ Value of the contingent claim: pÍp o xé o h œ ¿ “ ¥eÛ œ œ œ œ “ œ … œ i k i sQjh ” k ˜ µ Û q r˜ † p ¤k ” õ!n Ûo É n k ’ n ¤k n n ¢ ˜Y q  ‚ ’ ’ ‘ ¦™ ” ‚  ™ ” ™  q † ƒ‚ h n n h É n ‚ † ƒ‚ d q † k‚ q r˜ † n † ˜ k ” h  † p ¤k ” õ!n Ûo É n ¤k n h n k ’ ¢ ™ n  q  ‚ ’ ’ ‘ ¦™ ‚  ™ † ƒ‚ k ¿ Õn q r˜ ¶k ˜ CHAPTER 25. Û ui i jh ” ‰ p o p Then Intrinsic value: 25. either (a) or (b) is an equality. . p ut ” h Characterization of : Let p h 1o n is the smallest process with properties 1 and 2. Then is a supermartingale.6: Value of put with expiration be given.

. ÷ t If . . œ œ “ œ 262 p ÷ —Bo ” … is a supermartingale. 3.2. ” œ ui ۓ 1. Ÿ p o xé p o œ “ † p p o œ “ The optimal exercise time is ” d—(h o n ž œ  œ š ˜– • —bd†je† p p o ” !n 7(~ o ɇ œ œ ɇ … p p o ñé É7‡(~ o p p o o “ ” !n 7(~ … Q œ œ … i Õh ” p o “ ” !n 7(~ p o ɇ ” † is the smallest process with properties 1 and 2. then there is no optimal exercise time along the particular path .

). waiting until expiration to decide whether to exercise entails no loss of value.s.. so . be a convex function of . An Let American contingent claim paying if exercised at time does not need to be exercised before expiration.g. and assume .Chapter 26 Options on dividend-paying stocks 26. This stock pays no dividends.. we have . ‰p ˜ ëq k 1o † p ko ¤1xé ”p o œ h (*) Ê Æ ”† pp o ñé o p ho 1ñé k œ h “ p o ۉi iøh ” h  œ  œ o p o ‘Q™ Æ p o p o  p o † œ “ Theorem 1. (E.1 American option with convex payoff function œ œ p ¥£ £ £ £ ¤   where and are processes and a. i. For œ p ko é 51ñxù f p51oñxù p h o é p k é o ™ §1xñù q d ½i † p k ¤xù ™ xù d xé 5ñ1xé hp oo p k ùo q † h k Proof: For and .64 Consider the stock price process œ œ œ “ œ “  Æ i ù i d »Qjh ‘ p ko 51ñé  ¸ œ . The value of this claim at time Û Èi d i ¢ i ±h Æ p   5¡o  ß É ÿ sq þ í ýü ƒƒ¨† ú p¯õo Û ú i ûjh p o œ  “ © Û Let be the time of expiration of the contingent claim.e. 263 f ´p o œ Û pp Û o ¯Ço ñé “ pp Û o ͯõo xé p Û ¯õo d Consider a European contingent claim paying is ú ú p Û ¯õo “ p Û ¯õo ú i ³ p¯õo ú é pp¯Ûõo xé o Û p o p o ² f “ œ at time “ œ ú ° € p o œ ú p ÈÅ o † œ h ” ­ Û h p Û tõo and .

. ... ... .... .. . .. ... . ..264 Figure 26.... ... . . .. .. .... . . 26. .. ... .. .. .. . . ..e.. . . ....... ... ....... ... ... . . .. . . . ....... except in degenerate cases.. ... .. ..... .. ..... . .. ... the American claim should not be exercised prior to expiration. ... ............... .......... . .. i...... . . .. .. . . .... ...... ...... . .... . .. .. .. ......2 Dividend paying stock ¨ Let and be constant.. . ...... . . ....... . let be a “dividend coefficient” satisfying pp ko k ¤1xé ” 1o ¦ µ ¦ ¦§ k ¦ ” ui Û ¦ ¦ é ¦ œ ú ³ p o p œo ¥£ ú £ p Û ³ ´ p o ££ “ ¯õo ° œ p Û ¯õo œ ¥ ££ p Û ´ p o ££ ³ p¯Ço “ ¯õo Û p o ¥£ œ £ œ pp Û o ¯Ço “ ñé ¦ i jh ´p o §¦ § l f d ù œ ¦ ” p o oxé p QÅ p o  œ £ £ ¦ ¨ © ¦ œ l Õh ¦ “ “ ¦ ¦ p p o úf œ ú p o œ ú “ p o € ú œ ú ² õ° é ú p¯õo Û ú ° p o œ ¦ ¦ ¦ p QÅ o œ ú p o o xé œd ú † “ p o ² é œd ú † p o ² é œd ú  p o œ ú € d ú  p o p o p œ QÅ o ¦ € ¦ œd ¶ p ko é 51ñxù p k ùo 5ñ1xé † œ ‘ p p o xé o œ  “ ..... ......... .......... . ... .... ...... .. .1: Convex payoff function Therefore. the inequality is strict. .. . . . ... .. .. ...... ..... . ... . . .... ... In fact... ........ .. . . ...... (by (*)) (Jensen’s inequality) ( is a martingale) This shows that the value of the European contingent claim dominates the intrinsic value of the American claim. ...... ...... ........... .. ..... . ........... ... ........

p p ¢ o Proof: We only need to show that an American contingent claim with payoff need not be exercised before time . Options on dividend paying stocks 265 be the time of dividend payment. At times . the value of the call is f p p ¢ o Y¨ p œ “ o ” ¢ !n o q d œ ¢ At time . The stock Consider an American call on this stock. . the value of the American call is ’ ¢ p p o ” ¢ o Ñ(73É ~ ÐÉ ~4 ‡ 2 3É p ¤k ” o % Ï f œ œ … †„ ‚  e† € œ p p o ” o % “ œ œ “ 01kpY¨ o ” ¢ dn ” ‰ p o k 1o p 5k ” q d ˜ «q œ ( ý & • )Y'e† f œ ¢ o œ where % ” p ¢ o p œ “ ”¢ o œ % œ p 5k ” o œ † ¢ At time . According to Theorem 1.64. immediately before payment of the dividend. and let Let price is given by œ h ¿ Û ¢ œ .CHAPTER 26. the value of the call is . it suffices to prove œ “ ”¢ o œ % ” h  k ” 85Y¨ 0 p k p ” p o ” o n Ûui l ¢ p p mÄ o ”f ¢ i œ jhœ ” p œ o ” œ o ‚ i p þ “ ‚ % à† œ œ œ œ œ “ Ûui jh ” h h ” o % i p f œ † œ o ” ¢ dn ” ‰ p o k 1o p 5k ” ¢ o q d ˜ «q œ ( 7 6 & ¢ œ ý • Y e† œ % † % œ % (where ) with terminal condition ” h  k ”¢ œ i œ i jh ” h k ¢’ ‚ % ’ ’ ‘ ¦™ ‚ † ƒ‚ k  ™ ¢ % É % 5 i ™  »%  q i Õh Theorem 2.65 For . where f ´p # ˆ $’ ‘ ! "  o Vp œ q Ûõo ™ ˜ k  where f” ¢ Û Èi œ at time ” ui Û iœ œ l ¢ i jhœ œ l ¢ œ p tÛ ” ¢ o ¸ œ œ ” Ÿ pp ¢ o p ¢ o Y¨ d o p p o o p¢ o Vp ’ ’ ‘ ¢ q  o Ê q Ê ‘u™ q q p o œ œ œ ” Ÿ pœ o š í’ ý p ¤(ƒü o œ p h Ão ‘ Ê u™ ’ ‘ ¢ q  “ þ à† œ š í ý (ƒü “ “ ” Í5k ” pp œ q Û õo ~ uõ~ o ÐÉ Æ ßÏ ‡ (~ Û ‘ p ¤k ” Û õo ° q œ d  Æ q Þ † œ Û õo ‰ uk o  … œ ˜ q pp Í5k ” œ œ q Æ † p o ¤k ” !n œ œ œ p ho ¯Û ” 1e¸ ¢ be an expiration time. immediately after payment of the dividend. it is not optimal to exercise. so the value of the call is given by the usual Black-Scholes formula This function satisfies the usual Black-Scholes equation % and boundary condition The hedging portfolio is h p h ” ¢ o % 1.

The hedge has been constructed so the seller of the option has before the dividend payment at time . We k f p5kpY¨ d o ” ¢ !n o q œ 0 phpY¨ h o ” ¢ dn ” ‰ p o h 1o q d ˜ Âq œ † ( ý & • Y'e† k k p h ” ¢ o k ” o p5kpY¨ d o ¢ !n k q œ œ % p 5k ” ¢ o h §h ” ¢ !n p o † œ p 5k ” ¢ o % œ % 2. and the maximum of two convex functions is convex. If the option is not exercised. ‰p ˜ uq k 1o . and the seller of the option can pocket the difference and continue the hedge. is the number of shares of stock held by the hedge immediately after payment of the dividend. # of shares held when dividend is paid dividends received price per share when dividend is reinvested Case II: . Obviously. we need to show that is convex is . is convex in . @ 9 26. we have immediately that To prove that is convex in . The option need not be exercised at time have ¨ p ¢ o Y¨ p p¢œo p¢ œ o“ q d “ ™ p ¢ sÄ œ ‰p ‰p ˜ «q ˜ Âq ¢ o ¢ œo ‚% œ k 1o½lHp5kpY¨ o ” ¢ !n o q d œ k Ão † o p ¢ sÄ o q d o Ä sY¨ ™ œ † œ p ¢ sÄ ¨ q d o % †  † p¤kpY¨ o ” ¢ !n o q d p ¢ oœ p ¢ mÄ o œ œ d † ¢ p o ™ ¢ sÄ œ œ k Let .266 Since .3 Hedging at time œ “ † (should not be exercised if the inequality is strict). The post-dividend position can be achieved by reinvesting in stock the dividends received on the stock held in the hedge. its value drops from to . The proof of the convexity of in is left as a homework problem. p ˜ q k 1o p¤kpY¨ o ” ¢ dn o k q d ˜ kq œ ˜ Âq ¢ œ p sÄ o œ 4 A 3 B É É •–Þ ™ø† ™ p ¢ sÄ o p o ™ ¢ mÄ œ where ”p p¤kpY¨ o ” ¢ o Íp ¨ n o p 5k ” ™ ¢ osÄpY¨ q d o q ”d ‚ ” ¢q d † ” œ † p k 5pYœ ¨ o dn 5k o p q d œ † œ ¢ ¨ œ Case I: . œ is convex in . The owner of the option should exercise before the dividend payment at time and receive . Indeed.

which we call a stock. Define the accumulation factor I ) H p ¯Û ” o Given dollars at time . and we have already switched to the risk-neutral measure.Chapter 27 Bonds. According is ” 3o E Let be a Brownian motion (Wiener process) on some sider an asset. Assume that every martingale under can be represented as an integral with respect to . Con- . maturing at time . forward contracts and futures F p G” ¥ Here. its value at time and nothing before time . one can construct a portfolio of investment in the stock and money 267 Û f U V p o œ ¥£ £ £ £ ¢   ­ Æ Û” h ¸ Û © œ A zero-coupon bond. pays 1 at time to the risk-neutral pricing formula. which we call . whose price satisfies f p o œ C f Æ Q R  p o p o ‘ Q™ Æ p o p o  p o œ œ œ † œ œ ´ p o œ Æ p   53o “ € p¤3o É P    ÿ œ  ÿ q í ý üS € ß ¥ ££ þ ƒú õT e† p Û ´ p o ££ tõo ú ° p o œ ú  e† € p Û ¯Ço œ ú ¥£ £ £ £ É d ° p o í ýü (ƒë† œ € “ p o œ Û ú † p ¯Û ” o Ê œ “ Ÿ ei Û œ Æ œ i h ž m¯ƒp o H € ‘ œ ¥ ”p o C œ œ Ê  C D š . and are adapted processes.

you are given invest only in the money market. … ¥£ £ £ { pÛ ¯z Ço d ú ° y œ œ p o Ê   ú ú € œ ú † p QÅ o † œ œ † † † † p tÛ ” o p ¯Û ” o œ œ Ê Æ Ê œ œ   (*) . Any two of them are sufficient for hedging. then Û i   um©i f d † ” ¤3sÄ p  o p ¯Û ” õo Û Ê † p Ûo ¯ÇÈÅ p ¯Û ” o If.. at time . If. the money market. then is not zero. and the two which are most convenient can depend on the instrument being hedged. . at any time .268 martingale If is random for all . p tÛ ” o i. and the zero coupon bond.e. then at time you will have œ Ê ” f d ¢   † Æ p   53o ¢ œ ” p¯Û ” o p o p ¯Û ” o Æ Ê  Ê ¬Æ œ É œ œ † œ p ¯Û ” o Ê ÿ q þ (ƒë† í ýü œ   Æ  p   ¤3o ß  ß É ÿ Û þ í ý ƒƒü p ¯Û ” o œ … Ê œ Ä h † ¤… p o p o If is nonrandom for all . for some process . and we use the portfolio œ p sÄ o † œ We set œ Æ­ f p p œ o p o p o ‘ p sÄ ™ o p Q7p o o Å Æ Æ  œ o p C omÄ œ p œ oQÅ œ p o œ p o œ p sÄ † p ÈÅ oœ o Æ Æ q“  ™ © œ œ œ † œ “ “ The value of a portfolio satisfies . Indeed. then we will have dollars and you always p o œ C p o p o Æ œ … œ ú ú p o p o p o ™ p ¯Û ” o p o Æ Æ Ê f ´   œ pC   œ P … œ p ” œ h ° œ ú œ ™ Æ p¤3o Æ 53o É ÿ ™ ¯Û Ão Ê tp o p o œ C … œ œ ” ´ p5¡o   p¤3o P   p ” h ° É ÿ ™ ¯Û 1o Ê ¯p o Æ ú C … œ p Û ´ p¤3o ³ ¯õo   p53o P   ° ¯p o É ÿ ™ d ² € œ œ p o p o ‘ ú f pœ o p œo œ …“ œ œ C Æ … | ´p o œ Û market so that the portfolio value at time £ is 1 almost surely. One generally has three different instruments: the stock. Then given above is zero.

agreed upon at time . its value at time is p h Ão This suggests the following hedge of a short position in the forward contract. chosen so the forward contract has value zero at time . is not the value at time of is the price agreed upon at œ ´ p o œ f ú p¯Ço Û ú ° p h §1o p o ¥£ £€ œ ´ p o ££ 1o pp h p h 1o ¥£ œ £ £ £ Ð ú p¯õo Û ú ¤° p o œ´ p o Ðß P Ï Ï P Î 2 “ W œ † œ Y p h ¯Û ” Ão ¥£ † £ £ £ p o p œo € œ p Íp o p h Ão œ œ p ¯Û f p ¯Û ” o Ê œ ´ p o W aq œ p Û ¯õo W X ú p o p ¯Û ” o Ê p o f q ¥œ £ ú œ £ W p Û ´ p o œ ££ ¯Ço p Û tõo q W œ p Û ¯õo p Û ¯õo o p h ¯Û ” 1o Ê p h Ê §1o W W X q q Û “ Û ” o 1o p h Ê W Xq ú œ p o ph1o W Xq p œ o ¥£ ú £ p Û œ £ tõo £ ° “ p¯õo Û ú € p Û ¯õo “° “ o “ We solve for : f Û ui œ i ±h œ œ Û ” h † ´p o œ ¥£ £ £ £ p p o œ W X † d q p¯Û ” o p o Ê f p oœ † œ œ “ ú p o p œo ú † œ ° “ ú  e† € h “° p h ¯Û ” 1o p Û ¯õo o d “ € Ê p o œ ú † p o “ œ † ú p o œ † ú p o p o † œ  e† € p h 1o p Û ¯õo œ d W ú ° € ` p o œ p o œ ¥ W ” ­ Û h ­ ۔ h Û œ © © ¸ ¸ œ œ Û . Forward price) The -forward price the forward contract. This implies that W 27. In other words. this contract has value 0. At time 0. forward contracts and futures 269 27.1 (Value vs.e.. it does not. agree to pay for a share of stock at time . In fact. Bonds. however. The -forward price of the stock at time is the -measurable price. At later times. The value of the contract at time is zero. At time zero. time which will be paid for the stock at time .2 Hedging a forward contract Enter a forward contract at time 0. for purchase of a share of stock at time . i.1 Forward contracts We continue with the set-up for zero-coupon bonds. short -maturity zero-coupon bonds. This generates income Û W ´ p o œ ¥£ £ £ £ p o W œp o Remark 27.CHAPTER 27.

you have received the sequence of payments p ƒ o d œ p o d e œ  q p¢ ‰ o d e œ  pp oƒ p ¢ ‰ ƒ¡o o  q p ‰ o q œ¢ ƒ d eœ d i œ d e d œ b  œ ë† p o is -measurable. but the implementation of this hedge would require a term-structure model. The value at time of this sequence is Because it costs nothing to enter the future contract at time . Let us first consider the situation with discrete trading dates . the value of the future contract is maintained at zero at all times. This portfolio requires no initial investment. 27. this expression must be zero almost surely. either party can close out his/her position at any time. when the future price is .270 Buy one share of stock. you make the payment . ) The mechanism for receiving and making these payments is the margin account held by the broker. Thus. when the portfolio is worth A short position in the forward could also be hedged using the stock and money market. so f Û † b —Vœ l f f bVf p h 1o Deliver the share of stock and receive payment . Maintain this position until time . b at times . At time Enter a future contract at time . p¢ ~ œ o ƒ œ ¥£ ú £ å ú p¢ ‰ o £ p o £Rp oƒ p ¢ ‰ o go p  p o d u ‚Gc f è q c d 3œ ¢ ~ t á € œ œ c¡œ c 3œ ä b ” ” ’ ‰ ” ¢ ‰ P ff Vbf b œ œ † ë9œ d iœ d eœ ” p ¢ ‰ ƒ ƒ’ ‰ ƒ ” p ƒ p ¢ ‰ ƒ o p o o o q q q p b œ o ƒ ” ff VVf d i œ d e œ d i œ d e œ Û By time . when the future price is . is constant. € ¢ ‰ Vp o y wxp o q  c œ c œ c œ p   53o Q 1  Æ  4 ig h f P É ÿ dP u vc t sr p q I ) í ýü (ƒë† í ýü (ƒë† ú p¢ ‰ o d e œ p¢ ‰ ” On each . Through the device of marking to market. you receive a payment . taking the long position. f p h Ão W X q p Û tõo “ † p ¯Û ” õo Û l ¢ W œ Ê p h 1o l P œ ë† W X q h p Û ¯õo  c 3 “ œ c 3 #œ© d e œ Û ¢ ‰ ¥ e d œ . (If the price has fallen.3 Future contracts Future contracts are designed to remove the risk of default inherent in forward contracts.

then Û Èi œ i jh c ¡ œ ” h œ † U V p o ¥£ œ œ £ £ £ d œ p  o 5¡ƒ ¥£ € ú £ £ £ ¥£ £ £ £ p  o 53ƒ P É ÿ° ¥£ £ ptÛõo ° p ƒ o  e† œ € p  o ™ 53ƒ Æ ú p   ” p53oƒ 5¡o P   Æ d É ÿ † ú p   5¡o Æ p   ¤3o f œ “ Æ h d ”p o ” pœ o  ø† € p   5¡o œ ú † š  d p   ¤3o U ¡ ß ‡ ú d d p o Æ P p o Æ ú p œo œd œ ÿ ß ú É ÿ S T É ÿ ¥£ ß £ £ £ ¤ ¡   P S ƒ É p o  ø† œ € p o 5 ƒ ú ÿ S T € p o † œ p ƒ o Æ † œ p o ‡ Æ † œ p o € Æ ú œ p   ¤3o ‡ ‡ Û d ú ß ú p¢ ‰ o É ÿ S T p   53o c ¡ œ €  Æ Ð †  … ¢Ï É P „ . if (b) holds.s. forward contracts and futures The continuous-time version of this condition is 271 satisfying Theorem 3. -measurable.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.CHAPTER 27. If f Û Èi œ i jh ” ´ p o p o Definition 27. so £ Ûui jh ” h i U Vp o p  o £ 5¡ƒ Æ f œ † œ p¯õo Û p Ûo ¯õƒ † “ ” Ÿ ui j9ƒp ƒ Û¥ i h ž o is a martingale.66 The unique process satisfying (a) and (b) is On the other hand. then the martingale U V satisfies this implies p  o ¤3ƒ Æ p   53o d U e ú p o  œ P ´ É «q p o ÿ ¥£ £ ¤ ¡   £ £ p o 5 ƒ œ Æ ¥£ £ £ £ ¤ ¡   Û Èi jh i f ú ɜ p   53o p o 5 ƒ d p o ß Æ œ ÿ p   ¤3o S ˆ  Proof: We first show that (b) holds if and only if is also a martingale. as it should be when a. Bonds. and f (a) (b) is a martingale..

p Ûo tõƒ “ “ q p ho  §1ƒ3o † p  o ¤3ƒ Æ Û ß p Û ¯õo P Û ÿ “ W ‰ “ “ q p ho §1ƒ “ ß¢Ï Û . The cash flow received between times 0 and sums to Thus.  is a martingale. entered at time 0. he has paid a total of 27. .5 Forward-future spread ´p o £ Future price: f p h §1o W † p h 1o  p Û ¯õo Ð … If and are uncorrelated. the buyer agrees to pay The only payment is at time . Û ui œ i ±h £ ” ´ p o £¤p¯ÛÇo ° p ƒ o  e† œ € ¥£ £ œ Now define œ Æ term).272  and so is a martingale (its differential has no With a future contract.4 Cash flow from a future contract W Û p ho 1ƒ q p Û ¯õo † “ p Û ¯õo Û † p h 1o “  pp Û ™ ¯õo q  Clearly (a) is satisfied. p Û ¯Ço p Û ¯Ço “ Û f p ho 1ƒ p h 1o With a forward contract. so (b) is also satisfied. entered at time 0. if the future contract holder takes delivery at time . If he still holds the contract at time . for an asset valued at “ 27. the buyer receives a cash flow (which may at times be negative) between times 0 and . then he pays at time for an asset valued at . f ´ p Û ñ³ ¯õo p Û ¯õo ú “ ² € q pp Û Í¯õo o “ € p Û ³ ¯õo d ’Ð ß¢Ï … p h 1o „ ú ² “ € € ° q ­ Ð Ì ß¢Ï … Ë p Û ¯õo d “ Forward-future spread: f ´p o œ ¥£ £ £ £ Рߢ Ï … ° p o p o € œ ú † ptÛ ” o Ê p oœ p o † œ © € ª e† ¥£ Forward price: W œ “ œ “ œ € £ £ ptÛõo ° p ƒ o  e† œ € † p h 1o p Û ¯õo for an asset valued at . Indeed. By the tower property. is the only martingale satisfying (a). this f .

The owner of the future tends to receive income when the stock price rises.6 Backwardation and contango f p o œ C We have The expected future spot price of the stock under is ” ’ f g p Û ¯õo ‡ ~ Ð … Because is nonrandom. In short. the long position and . If the stock price falls. and p¯õo Û ¸ — ½™ò ” p Û tõo Æ f Ÿ Û ’ ’ p ÛH€ ’ ¢ q tõo ’ q ¥ p o ‘¥™ p o p o Æ Æ Æ  œ œ œ † œ “ “ ” “ H€ p — ¯3o ” – ˜¤ÿ † ¯õo H € p Û C š í ýü ¤(ƒë† p o – ’ ”•” ‡(~ ‘ ’ Define . Bonds. then . He withdraws from the money market to do this just as the interest rate rises. Therefore. Æ p o œ ‘ u™ Æ p o œ œ  ‘ pœo † œ C Æ Suppose p Û tõo “ Рߢ Ï f . 27. 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 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. p h §1o … “ W i p h H1o  “ “ ™ œ p o † œ C p Û ¯õo “ T † Ð C … If and “ ß¢Ï are positively correlated. and are uncorrelated under H € Ÿp o Ÿ p œo œ p h Ão f ß ’ p h … p Û Ão “ e† ¯õo ‘ ¥™ Û ’ ‘ ¢ q ߓ…“ C d í ý „ e(ƒü à € † € “ “ H€ p ¯Û ” o p o ÑÉ(~ ‡ Ð Ê f ß Ï e† p o œ œ … † œp o “ ¥£ “ W £ œ ” † £ p Û p ƒ o ­ p o £ ¯Ço œ © € † œ Ð “ p Û ¯õo ß¢Ï … ” ’ p h Ão “ ‘ u™ p ’ ‘ ¢ q  o C œ š ý p h í(ƒü Ão † ’ ‘u™ p ’ ‘ ¢ q o “ p o ú C œ  (ƒü š í ý É 7‡ † p œ o “ “ … e† œ f p o œ ” C ß … e† ß¢Ï ” Then is a Brownian motion under . the owner usually must make payments on the future contract.CHAPTER 27. but invests it at a declining interest rate.

This is called contango.274 l p Û ¯õo ¿ p ho —§1ƒ ¿ If then . then This situation is called normal backwardation (see Hull). . If .   f p h 1o ‡ p h 1o ß e† …  f “ h The future price at time is “ € p¯õo Û l h —p§1oƒ € “   .

i Qi Û œ i Õh Û – ” ho ¸ ­ Û 1»eÛ – Èi ±h ” Û i f œ Ÿ – ui j9ƒp o Û i h ž Suppose we are given an adapted interest rate process lation factor Q   œ ” ui Û œ E  š œ i jh H ” € Æ Ÿ – ¥i Û H p5¡o É P    ÿ ” p o p tÛ œ ” o € œ I ú Ÿ – Âi j—ƒp o Û i h ž i h ž Õ9ƒp o œ œ œ Ê p o í ýü ƒë† œ ú H Throughout this discussion. For . Therefore 275 ” Èi Û ”p o œ C œ ú ú p o p o p¯Û ” o 5™ i p Æ p¯Û œ ” o Æ ¯Û œ ” o Ê œ œ ú p o œ œ p ¯Û ” o ™ Ê ¬Æ œ d i jh ”p o œ C Æ œ p ¯Û ” o œ Ê p ¯Û ” o œ p o p ¯Û ” o ­ Âq  Ê œú œ j© ³ p o p ¯Û ” o Ê œd ² Æ œ p ¯Û ” o i œ p ” p ” ™ Æ ¯Û o Ê tÛ œ œ i 5 œ ú † ³ p o p¯Û œ ” o ² Æ Ê † œ p¯Û ” o œ  p ¯Û ” o Ê Æ o  ‘ † ” is a martingale under . We assume these bonds are default-free and pay $1 at maturity. equivalent to ).e.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.1 We shall always have for some functions and . . We define the accumu- œ œ C ¤ š € H † € p ¯Û ” o œ œ š Ê F p h” ¥ ” 3o E – Û . Theorem 0. .Chapter 28 Term-structure models C In a term-structure model.. such that for each . we take the zero-coupon bonds (“zeroes”) of various maturities to be the primitive assets. the process Remark 28. let price at time of the zero-coupon bond paying $1 at time . and W is a Brownian motion on some probability space is the filtration generated by .

Option expires at time œ f – u½ui Û i Û b H € p o œ œ ff VVf H € i jh C Æ p o p tÛ ” o Ê  œ ú œ p o p o p ” p ” ™ Æ ¯Û o Ê p o  ¯Û o Æ Ê œ … œ œ œ œ † œ Ÿ – ¦i Û ” Vp o U œ ¥£ £ £ £ i h ž «‘ƒp o Now that we have an interest rate process prices to be £ . we define the zero-coupon bond p o œ €  œ p ¯Û ” o n p o  œ œ Ê p o œ C  f p o œ C p QÅ o H œ Æ € p o Íp QÅ ” o œ ptÛ ” o H € p Ê œ ¯Û ” o n œ œ p o m 5 œ ¢ p o ” ™ Æ p QÅ o œ œ œ    ” Æ b H ´p o p   ¤3o œ œ œ f f bVf    ¥ £ sq ÿ í ý üS € ß ú þƒ(ƒÑƒ e† £ ú p¯õo Û ° p o p ¯Û ” o š £ É Ê o H d p ¢ õo Û k l d H p QÅ o Æ † œ n ú ° € œ € p o œ Å † ú œ œ p o œ p QÅ o œ … Ê  Û p QÅ o œ H € . we should change to a measure under which the mean rate of return is . Cox-Ingersoll-Ross. is a risk-neutral measure 28. Since and there is no arbitrage. We showed in Chapter 27 that for some process . we take to be (e. the mean rate of return of under . We let the interest rate of . we let be an -dimensional be a function Brownian motion and let be an -dimensional process. has mean rate of return under ..g. H 28. If we want to have -factors. If such a measure does not exist. Price at time is .1 Computing arbitrage-free bond prices: first method Begin with a stochastic differential equation (SDE) The solution is the factor.2 Some interest-rate dependent assets ۔ p ۔ o Ê f d œ ” “Û ” ¢ Û ’ d RH Û l ¢ Û f ¢ ui Û œ i ±h ” ´p o Û Call option on a zero-coupon bond: Bond matures at time Price at time is œ ¥£ £ £ £ ‰p ˜ «q  p ¯Û ” ¢ õo Û f ß ” Ž 1 ôÉ t d Œ ” ’ ”¢ Coupon-paying bond: Payments at times . 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 . then the model admits an arbitrage by trading in zero-coupon bonds. HullWhite). .276 . In the usual one-factor models.

1 (Term-structure model) Any mathematical model which determines. you receive $1 from the -maturity zero. Suppose you want to borrow $1 at time with repayment .3 Terminology Definition 28. Y p f tÛ ” o Ê œ p r At time .1) œ i jh q œ Û õo ” tÛ ” o p i uh š í ý t(ƒü œ p tÛ ” o Ê œ œ œ Ê œ p © – ¥i ™ Û p 5 Û Û l Û ™ ¥À¥i – ” ho ¸ ­ Û 1Â½Û Û œ i Õh p o œ ¥ . To synthesize a forward-rate (plus interest) at time agreement to do this. The (4. at time buy a -maturity zero and short -maturity zeroes. 28. the yield to maturity œ ” ui Û is the . The value of this portfolio at time is f p ¯Û ” o œ Ê   xo Þ Û u ˆ u q † p ۔ ۔ o P p tÛ ” o p©™ œ ¹ ™ë† •–Þ œ A q t The forward rate is    q p p p s   xo q † p p 5 ” ” ™ Û Û o œ ¹ or equivalently.CHAPTER 28. p ¯Û ” sÔ o ” d † Ÿ tÛ ” o €p p Ô – Û œ i Û œ i Definition 28. you pay $ given by Ð ‚‰ 3É 2 q Ð 3ß É Ï 2 ß Ï Y p Û p q Û ™ õo p p5™ Û ” Û ” o ™ Û œ ¹ Ð ‚‰ 3É 2 q Ð 3ß É Ï 2 ß Ï Y f ” Ÿp ۔ ۔ o p5™ œ ¹ h œ † Y p f Þ p 5 – uÀÈi Û i Û p 5™ Û ” o ” p p ” ™ Û o Ê ¯Û ” o œ Ê q ¯Û o Ê p Ê œ œ ” ™ Û o Ê œ Û œ is equivalent to determining ” – uÀÈi Û i Û œ p š í ýü ¤ƒë† i jh i jh œ Þ ” ¯Û ” ‡Ô p o œ ” ¯Û ” o p Ê p p©™ Û ” o Ê p ¯Û ” o œ Ê œ Determining f p ¯Û ” o œ Ê   xo Þ q œ d Û q † p tÛ ” ‡Ô o Û Û œ or equivalently.2 (Yield to maturity) For -measurable random-variable satisfying . at an interest rate agreed upon at time . At time effective interest rate on the dollar you receive at time is . Term-structure models 277 28. at least theoretically.4 Forward rate agreement Let be given. the stochastic processes for all .

u ¥£ £ £ £ ¢   ” Up o œ Æ p   ¤3o ¥£ £ £ £ ´p o p o f  q œ É † œ  ß   ¢ ÿ q Æ p   ¤3o  £ ° £ yp o   q œ p Û ¯õo ” Vp o U œ þ ƒü íÉ ý ß ÿ sq   q S þ í ý üS € ƒ(ƒÑˆ e† x 9 w ¤$"v É ££ Û ß£ £ p ¤¯Û ” o u Ê u  €e† œ Û u p¯Û ” o Ê u  €e† œ p ¯Û ” o Ê œ É ß t d † Û p ¯Û ” o œ Ê ” ­ Û ¸     ¢ Æ p ¤  ” o œ   ÿ þ í ý ï(ƒü p tÛ ” o p 5  ” o œ œ Ê 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 . agreed upon at time . Û œ Conclusion: at time .”   ¢ Æ p ¤  ” o œ ß t   ¢ É ÿ sq Æ p ¤  ” o œ þ í ý ‘ƒÉü t p ” o f œ œ p ¯Û ” o t œ ” t ß ¥£ £ ÿ sq þ í ýü ƒ(ƒë† p ” o p o  É ££ œ œ † œ £ ß ¤¯Û ” o t Û u £ p u Ê u q † œ Û u p¯Û ” o Ê u q † œ p ¯Û ” o Ê œ . for money borrowed at time . we obtain 278 28. On the other hand. so Integrating the above equation. this will grow to œ p ¤  ” o œ #œ© É t œ f ¢   Æ p 5  ” o œ t ß ÿ sq t þ ƒë† í ýü p tÛ ” o œ Ê É u † u †£ £ ” p¯Û ” o £ Ê œ p 5” o £ ¤     xo œ Ê ß Þ q q † †   Æ ß p 5  ” o œ   u u   xo É Þ É t Ê   xo Þ ÿ q †   Æ p ¤  ” o œ ß ÿ This is the instantaneous interest rate.5 Recovering the interest Û œ from the forward rate .

Term-structure models For each .6 Computing arbitrage-free bond prices: Heath-Jarrow-Morton method CHAPTER 28. f p o œ C p¤  ” o Æ œ p ¯Û ” o Æ œ and Then Let Recall that t ß ÿ q ~ } Æ p ¯Û ” o œ Ê ¬Æ are adapted processes.œ p C Æ ’ ’ pp ¯Û ” o – œ œ Æ ’ p– – q œ ™ o Ê † Æ – ¡ ~ } ‘ q Æ œ † † p o p ¯Û ” o ¯Û ” o – ‘ p Æ Ê f œ p ” ’ œ pC ” – œ ‘ o ¢¦™ ¯Û o ù p o tÛ Éq œ  „ œ ” € ¡¡ ’ ‘ o  p5  o Æ ÿ sq ~ } ¢ œ ß É t ù o €   5  ” o p Æ q Æ  ÿ q œ œ ßÉ p 5  ” o t €   Æ œ ” É t €   Æ p 5  ” o œ ß ÿ q p ¯Û ” o † œ Ê ~ } | C p o œ Æ U |  É t p¤1Ç ¡¡ ” ko p¤k1Ç¡ ” o p k 5Ão f } } } … … ‚ …e† ‚ e† ‚ e† p o p ¯Û ” o – p ” – ‘ q Æ ¯Û o ù q Æ p o  Æ Ð ¡É Ó f 2 Ð œ 3É Ó 2 œ œ œ † ß Ï ‘œ C ß Ï { œ É y É y {z | {z p ” p¤  ” o ‘ ÿ q Æ Uz  Æ 5  oHù ÿ q Æ p o  Æ œ œ œ œ œ † ß S ß ÉS ”   p o p5  ” o p o ‘ u™ Æ p¤  o ù Æ ­ Æ ÿÂq Æ  œ C œ œ œ ©É ß œ œ †   p¤  ” o p ” o Æ Æ ÿ q Æ œ œ † œ t   ¢ ß œ Æ p ¤  ” o œ ß ÿ q þ Æ Now t f ¢   f Û Èi œ i jh ” 53o p   C Æ p ¯Û ” 3o ‘ É P ™   ÿ É p ¯Û ” o Ê ÿ mq þ í ýü ƒ(ƒë† œ ß t p ” p ” ‘ u™ Æ tÛ o ù tÛ o Æ œ œ † œ t Ÿ Ûui¤ siÕxž ¯Û ” 3o h p   Û i   i h žp Ès©jñƒ¯Û ” 3Hù   o ‘ š š p ¯Û ” 3Hù É P ™ ¯Û ” 1o   o p h ÿ   Æ p ¯Û ” o † œ – ” ho t¸ ­ Û 1Â½Û 28. 279 . let the forward rate be given by t Here and Ÿ In other words.

we must have and . but there might still be a riskbe an adapted process.r.. Let (7. let for all i jh – o ’ ‘ ¢ H€ Û Èi ±h i f ’ p o(’GtÛ ” o – p pp œ ” – ‘¥™ ’ tÛ o Æ œ œ œ œ ’ pp ” – ’ tÛ o œ œ Æ ” p o(G¯Û ” o – ’ p pp ” ‘ ¥™ ’ ¯Û o œ ” œ p œo  ” p o œ ” tÛ ” o ¯Û p p Æ Ê ’ œ C ” – œ ‘ o ¦™ ¯Û o ù p o ¢ p q œtÛ ” o ¯Û p o p pœ Æ Ê ’ œ C ” – œ ‘ o ¦™ ¯Û o ù q p o ¢ p œ œ under ¥  ” o – ‘ q œ ¯Û ” o p Ê  † ”„ o – œ ‘ q œ ¯Û ” o p p ¯Û ” o „ œ Ê † œ œ Ê Ê Æ Then p ” Q   Æ f p   53o ’ ’ É P ¢’ ÿ Not only does (7. be a deterministic function.3) (7. f – u½ui Û i Û œ i jh ”   É H € Æ p 5  ” o ‘ œ ß ÿ É ß ÿ p ¯Û ” o ‘ œ †   f – u½ui Û i Û œ i jh ” ’ €   Æ p 5  ” o ‘ œ ~ ¢ ’ † ” Û i Û Y– Èeui œ i jh ” p¯Û ” o – o ’ p ‘ ¢ ’ œ † Suppose (7. and define Û – t Âi jh ” ¯Û ” 1o Û i p h Û i  œ i ûrƒeh ” ¯Û ” 3o ‘ p   t . (7. and assume . This will be a homework problem.1) imply (7. Then neutral measure. yields the equivalent condition œ Theorem 7.1) does not hold.2) also implies (7. be adapted processes.2).1) ” p o(’G¯Û ” o ” – p ” p ‘ p ¥™ tÛ o ‘ ¯Û o ‘ œ œ œ œ p tÛ ” o ù p ¯Û ” o H € .7 Checking for absence of arbitrage 280 is a risk-neutral measure if and only if to have mean rate of return † † ۜ p tÛ ” o – ù œ Differentiation w.t.1). and define . Û œ œ Æ p o 5  ” Hù ß É ÿ i.t.68 (Heath-Jarrow-Morton) For each .4) (7. Let – õo Û ¸½™ò — p – õo Û p — ¯¡o ” – ˜ ÿ Æ H€ † H€ p¤3o   p¤3(’ É P   o p o – Æ ÿ q q I (ƒë† C í ýü œ P ”p o ”   o ™   Æ p53(’ É ÿ p o † œ C i h ž o ±—ƒp ’ œ C Ÿ – ui Û œ œ š p ¯Û ” Hù o Differentiating this w.f and p   ¤3o C Æ p¯Û ” 3o ‘ É P ™   ÿ   Æ ptÛ ” ¡o ù É P ™ ¯Û ” Ão   p h ÿ † t p ¯Û ” o œ Û   h ¿»¯Û ” 3o ‘ p   – ” ho ” ۉi  ‚iëh ” p¯Û ” 3oHù   ­ Û 1r¸ i jh f – uÀ»i Û i Û – uÀÈi Û i Û f œ .2) (7. p ¯Û ” o – ù œ In order for 28. we obtain is not a risk-neutral measure.e.r.

for every maturity : 28. satisfying (7. so we can solve (7. satisfying (7. (The remainder of this chapter was taught Mar 21) Suppose the market price of risk does not depend on the maturity . The excess mean rate of return.4) is equivalent to (7. we obtain for every maturity : ’ Because (7. the zero-coupon bond with maturity has mean rate of return – ½i Û œ i sh ” p ’ o œ p ¯Û ” o – p ” p ” ‘ q Æ ¯Û o Ê p o  ¯Û o ¬Æ Ê Ê œ œ œ œ † œ ” o Ê œ œ œ œ œ œ œ ” p ’ o ” ¯Û ” o – pœ ‘ œ S œ q Û p (’ o † œ – ½i Û œ œo o ‘ ‘ © † † ¯Û ” o p Û ½i H € p ¯Û ” o – ‘ œ œ t Æ œ i gh ” ¯Û ” o p œ p ¯Û ” o t œ t Û .8 Implementation of the Heath-Jarrow-Morton model Choose Û p o œ f C p o Æ œ p o ” ¯Û ” o u™ p p ¯Û ” o – ‘ ¯Û ” p ‘ Æ Æ f œ p ¯Û ” o Q™ œ C p (G¯œ Û ” o u™ ¯Û ” œo – ‘ ¯Û ” o ’p p ‘ ‘ p Æ ­ ” C p¯Û ” o Æ œ p ¯Û ” œ Û U p ¯Û ” o – ‘ p p ” – ¢’ œ p ” – ’ tÛ o ‘ o £™ ¯Û o ù q œ ’ pp¯Û ” o – ’ p ” – ‘ o £™ tÛ o ù q p o  ¢ œ œ – ui jh Û i ” – uf Àui œ jh Û i Û i p ¯Û ” o – ‘ ” – f pp ” – ’ ¢ œ p ’ tÛ o ‘ o £™ ¯Û o ù q œ œ œ ” pp ” – ’ p ” – ’ tÛ o ‘ o ¢ ™ ¯Û o ù q œ œ Û Remark 28. and volatility .3).3) for .CHAPTER 28.3). or equivalently.4). we may plug (7. Plugging this into the stochastic differential equation for . 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.4) into the stochastic differential equation for to obtain. 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 .2 Under . is and when normalized by the volatility.3) is satisfied. above the interest rate. We can then set and (7.

the volatility in œ Ê f – ueujh Û i Û i ” h † œ p ¯Û ” õo – ‘ Û d œ obtained from . it is apparent that neither nor will enter subsequent computations. gotten from the market. but are usually taken to be deterministic functions. it suffices to have the initial market data and the volatilities p ¯Û ” o Ê f – ueui Û i Û Û œ i Õh ” p¯Û ” o – ‘ œ p ¯Û ” õo Û This is because vanish.1) f – ueujh Û i Û i ” ¯Û ” 1o p h Ê   xo Þ Û u u q † p h ¯Û ” 1o t ” – u½ujh ” tÛ ” 1o Û i Û i p h Let be determined by the market.282 These may be stochastic processes. combined with the stochastic differential equation H € Ê f p o œ p ¯Û ” œ É ” – ÛuieÛÈi ±h ”   5  ” o i p p ” ‘ ÿ ¯Û o Æ œ œ † œ ß – Û i Û ueui ±h ” ¯Û ” o ‘ i p p ¯Û ” Hù o ” C Æ p ¯Û ” o œ ” – ui Û Ê p ¯Û ” o – ‘ p ¯Û ” o p o Ê  q Æ œ œ œ i jh p œ (’ o œ œ ”p ” o œ œ t this determines the interest rate process ”p o œ ” C Æ p ¯Û ” o Q™ p ¯Û ” o – ‘ tÛ ” o ‘ p ‘ Æ œ œ œ œ p o  † œ œ œ œ † p ¯Û ” o † p ¯Û ” o œ – œ o – ‘ Æ Ût ui ‘ † Ê ¬Æ p ¯Û ” o ‘ œ i Õh Ê ” C p ¯Û ” o Then for is determined by the equation (8. we have written (8. Equation (8.3) we see that .1): ” Q   Æ f p   53o ’ ’ É P ÿ ”p p (8. . recall from equation (4. under which is a Brownian motion. and the process The only process which matters is (8.2) (8.4) implies i Õh ” ¯Û ” 1o p h In conclusion. Define Because all pricing of interest rate dependent assets will be done under the risk-neutral measure . is the volatility at time of the zero coupon bond maturing at time (8.3) in terms of rather than .1) and (8. to implement the HJM model.5) must i Àmh ” ¯Û ” 1o Û i p h and then the zero-coupon bond prices are determined by the initial conditions . and so as approaches (from below).4) ” ” C From (8.3) œ – õo Û ¸ — ½™ò H€ ’   p  o ¢ q p53o Æ 53(’ C ”p o ™ œ C o(Gp¯Û ” o Q™ ¯Û ” ’ ‘ p ¥ œ œ Æ É sq ÿ I (ƒë† í ýü   p  o 5¡’ P É ÿ Æ † o – p ” ‘ ¯Û o ‘ œ † P p – õo Û – ˜ p — t3o ” ÿ † H€ p o – œ p o ” œ C p ¯Û ” Hù o œ œ t t œ ” – »ÀÛ Û i – Û C Û . Written this way.

In (8. and volatility is unaffected by the change of measure. The only parameter which must be estimated from the market is the bond volatility .4) will be satisfied because Remark 28. Term-structure models 283 and satisfy (8.3 It is customary in the literature to write rather than and rather than .5). where is given by (8. and we let by (8.1). We can then define and (8.3).1) we use the œ œ Ê H € f Æ   C f p¯Û ” o ” œ H€ p 5  ” o – ‘ – ueujh Û i Û i œ   u t u ” ptÛ ” o – Û ‘ u ß É ÿ ” ¯Û ” 1o p h u œ p ” o – p ” – p ” – ‘ q ¯Û o ‘ ¯Û o ‘ † œ œ œ † œ Ê †   xo Û We require that p ¯Û ” o ‘ Þ Û u u œ p o q œ  † p h ¯Û ” 1o t p ¯Û ” o – ‘ œ be differentiable in ” C € – ½i Û œ p ¯Û ” o – ‘ Û .2) and initial conditions C . is the symbol used for the risk-neutral measure and no reference is ever made to the so that market measure. be given by (8.CHAPTER 28. ” ” H i H € i Õh ” ¯Û ” o p We then let be a Brownian motion under a probability measure .

284 .

. 285   ” ” ¤  ” ¢ ’   .1 If is jointly normally distributed. is a stochastic process with . . then its distribution is determined by its mean function p oQÅ p o  e† œ € œ „ and its covariance function p o ‰  ’ ” k ¢ k f"•ippV’o ‡ ˆ   o ’¬† ‘™~ † pVŽo p „  Žo ’ ¢ Œ Š ™‹ü Æ fVVf Æ q f “”q q ˆ d Db b Œ d eí ý Ÿ k Hp QÅ (ƒü ” ¢ k Hp ¢ QÅ † ¸ o ” ¸ o Æ Æ f f bVf b œ š H€ p QÅ ” b œ ” p ¢ QÅ o o ff VVf b œ p oÈÅǬ†RpV…3o o p pV3oQÅBo … p œ ” 3o i … © € ª e† œ p Íp o f­ œ „ q œ „ q Å where Indeed. is a Gaussian process.1 (Gaussian Process) A Gaussian process the property that for every set of times Gaussian processes Chapter 29 . the joint density of is the covariance matrix ” ”’ ”¢ ­ f f bVf œ ©œ ƒ’ ” o p b œ” o p i æ ff Vbf œ bœ ææ åè b fœ f b œ ff p V” b’ f o i Vbf ƒ’ Vbf o p f”f’ ff Vbf p œ ” ¢ œ o i Vbf p ’ œ ” ¢ œ o ff œ œ œ b  œ .” f ¢   d b ­ b f f bVf © s— † • — † Œ † — ’ ¢ ¦™ • p V’o „ † — d ‚e p QÅ o ¢ d e í ýü (ƒë† œ u –d t þ ƒü € í ý is the row vector ­ b œ ff Vbf œ œ © † b Œ  tœ œ Definition 29. the set of random variables h ” ’ ÈÅ ” p ¢ QÅ p QÅ ” o p o o f f bVf b œ i i ¯’ œ i ¢ jh iœ p b QÅ VVf oœ f f œ œ Remark 29. and p o „ ” ” D’ o p „ ”p¢ o „ p VŽo „  i i i ­ p¢ ” o ib fœ ” f ’ b œ p ¢ bVf o i ãä ã p¢ œ ” ¢ œ o i á㠜 œ The moment generating function is k” ff VVf ” ¤k ” ¢ k ’ © † uŒ where  . is the row vector is .

Then ¢   þ í ýü ƒ(ƒë† Ÿ p ÈÅ ’ o   œ C C o … f Ü † p… V3o ’ œ p … C € b3o ˜ e† f € V3o p… C C j „  e† € p o o q œ C ­p o œ C € ˜ e† p ” 3o … © € Q e† œ i Proof: (Sketch. and shall cheat a bit by considering only two times.69 (Integral w. if i … © i jh Theorem 1.f.1) . Indeed. a nonran(1. Æ . is a Gaussian process with is normal with mean 0 and variance and Martingale n ¨ šP p …o b3QÅ be a Brownian motion and .Æ p n ¬1o ’ „ Ð š Ï  t‹† š Ï  t‹† Ð Ð Ð … j € … … j € Æ … j € š Ï  t‹† P ’ Q ‚n p n ¬1o ’ ¨ Ÿÿ ’   ¢ Æ f š I í ý ü )(ƒë† P ’ Q n Ð p n 1o ’ ¨ Ÿÿ ’   ¢ Ï  Æ š P t‹† I ƒü í ý … e† ”К ’ p… V¡o ’ ¨’   ¢ Ï  ¯‹† … j € † P ” n Ð ’ p n ¬1o ¢ ™ ’ ¨ Ÿÿ ’   ¦wd žÏ  Æ t‹† š … j € † | { Dz y P P ” n p¬1o n   ’ n p no ¨ ¬ÃVÐ   ™ Ð Ï  ¢ ™ p¬Ão ’ ¨VÐ Ït‹† ž  žÏ  o¯‹† Æ Æ š ÿ ’ š ÿ P t‹† … ” … pV3o Ð C … … e† …   ’ p … b3o p …o ¨ b3(Ð   š Ï  t‹† ’ ¢ ™ Ï  t‹† Æ … C … † Æ ’ ¨ š š Ït‹†  … Ð Ï  t‹† Æ … š f Æ p   ¤3o   p¤3o ’ ¨ É   P œ š › ÿ p ” 3o … † œ i C Æ œ p5¡¨ É P   o ÿ p o C p QÅ o † œ h p o † œ „ f ¢ U p ¨ o œ ’ 5  ¢   S U ’Í’ ‘ ’¢ ‘ ¢’ ‘ ¢ ¢ ‘ ­ ’ S   ¢   © ’ ¢ ™ ’ „ ’   … ™ ¢ „ To prove that a process is Gaussian.g. a Brownian) Let dom function.t.. which we usually call and . We shall use the m.r.1 An example: Brownian Motion 286 Brownian motion then is a Gaussian process with and .) We have f C Æ ¨ † Å Æ p …o ™ V3QÅ ¢   This shows that š í ýü t(ƒ» € Therefore. one must show that has either a density or a moment generating function of the appropriate form. We will want to show that p o ÈÅ ” b  œ ff Vbf œ ” p ¢ QÅ o œ ’ pb3o ’ … C pp… ™ V3o Ü … pb3o … pp… ™ V3o ’ C p ™ € xV¡o o o p… C i q œ C œ œ p ” ¡o … † œ h p o † œ „ C 29.

Gaussian processes be given. Ï t … ° € (1.f.2 ´ V¡o p… ¥£ £ Ð ÑÉ Ï ¢ te‹† ‰Ð š Ð ÑÉ Ï  t‹† … Ð ÑÉ ° € ° … … ¥£ £ £ £ ¥£ £ £ £ Ï  t‹† Ï  t‹† œ ° € € Æ Ð ÑÉ ¥£ £ £ £ ¤ p n ¬Ão C Æ Ð ž  Ï  t‹† … p… ­ V¡o Ï  t‹† … ™ Ð š t‹† Ï  Ð ÑÉ Ï  t‹† … … e† Æ Ð ÑÉ œ This shows that pb3o Å … ’ € p o p ÈÅ ” V¡ÈÇo p …o Å The solution to this ordinary differential equation with initial time is š ¥ p¬nͨ É ° o ÿ €   ¡ f f bVf © € … l … i ƒsÕh to get Take Integrate from to to get conditional expectations and use the martingale property is jointly normal with .P † œ ’ Ð ÑÉ Ï ¢ te‹† ¥£ £ £ £ ‰Ð š Ï t 4 † …„ †¦ € 4 † f … ”  œ f … ¥£ £ £ £ œ Q ™ n Æ … i ©±h … i œ š p1o ’ ¨ É ’   ¢’ n ÿ : ¥£ £ Ð š I ) í ý ƒü Ï  t‹† … e† ´ b3o p … £ £  Hœ Ð ž o n Æ ´ V3o p… Ï  t‹† ” ´ ´ V¡o ° p… pb…3o ££ ÑÉ Ð p o ’ ¨ ’   ¢’ Ï  t‹† € œ † š … ´ V¡o ’ ° p1o n p… ¢ ’ ¨ É ÿ ’   ¦™ Ð š t‹† Ï  … ë† … € ¥£ p n 1o C Æ Ð ž o Ï  t‹† … p no 1(¨ P š Ÿ £ £ £ ¤ h p n ¬Ão C ÿ q ´ V¡o p… š Æ Ð ž o Ï  t‹† … Æ C pn1(¨ É P ° o ÿ Ð † ´ p… V¡o  ë† € š n Ð p n 1o ’ ¨ É ’   ¢’ ™ ¬Ão p n žÏ  f Æ ot‹† ÿ … ž o Ï  t‹† … Æ pn1(¨ É o ÿ   p ¨ ÑÉ o Ð œ œ Now let CHAPTER 29.g.1) 1. ’ p o Ð ¢ ’ ¨ ÑÉ ¯‹† ’   ¦™ p o Ï  f Æ œ œ … œ C Ï  ¯‹† …   † We now compute the m. for n ¬1o p n o Åp …o ’ ¨ Ÿÿ ­ p Q7V3QÅ f Æ š † œ © € ” n 1o ¨ P ” n 1o ¨ P p n p o Å p n ’ É ÿ ’ ’ Ÿÿ Æ Æ š † œ € h p QÅ o p …o V3QÅ † œ 5   e† É € ¨ € ’ ’ P ’ ¨ š P „ 5  ¢   ’ U ’ ¢ ¢   U ¨ šP „ ­ f ’ ’ ¨ šP „ „ S S © ƒ(ƒë† þ í ýü ¢ P P ¢   ’ ’ ’ Q‚n p n 1o p n p’ ¢ ’ ¨ É ÿ ’   ¢ ™ n Æ 1o ’ ¨ £ÿ ƒ¤    ˆ ™ ’ ¡o ¢ Æ š I í ý ü )(ƒë† š ’   ’ Q ‚n p¬1o ¨ n n n1o ¨ P ƒ¤  p p’   ’ ™ ¢ 3o ¢ ’ É ÿ ’ ¢ ™ Æ ’ £ÿ ’ Æ š I í ý )(ƒü † š ’ ’ Q‚n p n ¬1o f iÐ Ð ‰4 ’ ¨ É ÿ ’   ¢ ¥£ Ï ¢ † Æ f š t ‹† oÏ I7(ƒü … d € 5s e† £ í ý £ Q ´ pV3o … £ $ÐÑÉ ‰Ð 4 ° ÏtG‹† š Ït †  ¢ … € I  e† € š ” Q‚n n1o ’ ’ Ð Ð ‰4 p š ’ ¨ É ÿ ’   ¢ ¥£ Ït ‹† oÏ Ð Ï ¢ † Æ I)(ýƒü … † £ í ´ V3o ££ ÐÑÉ ° Ð 4 p… š Ï ¢ ¯i‹† Ï ¯ † € … e† p o p QÅ ” b3Qõo p …o Å … (1. where . Just as before.2) 287 Æ .

f h  Hœ ò h p ÈÅ o p o † œ  ø† œ € „ h p ho §1QÅ is a martingale and . so p   ¤3o C Æ p5¡¨ É P   o ÿ p QÅ o † œ †  p …o V3QÅ ¨ . the Itˆ isometry says o However. n Æ p¬Ão ’ n ¨ š Ÿ P ÿ † pb3o ’ Å … € € œ Å ¨ p QÅ o œ ¨ h … For fixed . Therefore.3 When is nonrandom. We proved this before. The equation (1. p QÅ o œ n Æ n 1o ’ p n f Æ p¬Ão ’ ¨ É š n „ ¨ p   ¤3o € p … b3o š ¥ P i ÿ ¥ … © p5¡¨ P   o o É ÿ p QÅ Æ † œ † pV3o ’ Å … i jh and the same argument used above shows that for . For o . is to prove the normality. and the reason we use moment generating functions. The computation of means and variances does not require the use of moment generating functions. is normal with mean and variance .2 The hard part of the above argument. so its distribution is not determined from its mean and covariance functions. P n ¬Ão p n p … b3o ’ ¨ ’ Å f Æ š £ÿ  ø† o Å p … o € p … pp …o o †Å o p … o ­ b3o ’ Å ™ ÍV3QÅ q p QBVb3QÅ ­ p Q7b3QÅ œ ©  ø† œ € © € If were a stochastic proess. the distribution of depends only on . when is stochastic.2). but note again that the Markov property follows immediately from (1. æ | ææ åè ³ V3QÅ p …o ´ ¤V3o Q p … ¥£ £ £ £ q pp …o b3QÅ { z ´ b3o p … P ¥£ £ £ £ q p ÈÅ ° o œ p Q™VV3QÅ o Åo p …o n œ Æ œ p n ¬1o ’ € ¨ y p …o ² V¡ÈÅ ãä € I q š P p …o V¡ÈÅ ÿ € áãã ° †  e† € f  ø† €  ø† € p … b3o ’ Å h † o Åp …o ­ p Q7b3QÅ œ © € … o Åo p …o ­ pÍpV3oQÅ q p QBVb3QÅ œ © € i … i ƒsjh by the Itˆ isometry. in fact. Indeed. C is also Markov.288 Remark 29. is not necessarily a Gaussian process.2) says that conditioned on . Remark 29.

Here we observe that and we verify that (1. n ³ ¦ ¬¡xé É ž ² ³ ¦ 3xé p ¦o p ¦o Æ Æ f Æ ÿ ” h †   Æ p  o 53QÅ œ i € p¤3xé É P   o ÿ … T ¨ ©„ i Âh š ž p n 1o ’ ¨ É ÿ ² p ‡Ô o p o † œ  e† œ € p ” 3o … œ ¨ zi œ Vš P ÿ p ” 3o … † œ ¨ ©„ p o p …o Ô p ªÔ ” V3sõo œ ¨ i Ô Then is a Gaussian process with mean function and covariance function f Æ   p xé o p5¡oÈŀp53oñé É P     ÿ p (¨ o h p o † œ p ¤Ô o † œ ” 5¡o p   C Æ p53(¨ É P   o ÿ p ÈÅ o † œ p o Theorem 1.CHAPTER 29.1: Range of values of for the integrals in the proof of Theorem 1. z œ C be nonrandom functions.70.3) . and let and œ œ ” § n (–” ¦ Figure 29. Gaussian processes y= v= 289 z z z z s y t (a) y= s v (b) z 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. (1.70 Let Define be a Brownian motion.3) holds.

4 Unlike the process We have 290 .1(a)) is p n 1o ’ ÿ ² ¨ P š › ¨ š ¨ P š › P ÿ † ÿ † ÿ † p n 1o ’ ÿ ² p n 1o ’ ÿ ² P š £ ÿ ™ ¨ P š › p n 1o ’ ÿ ² ž P š ¥ £ š ž ÿ † ÿ ÿ ™ P š Ÿ › š P ÿ ÿ † ÿ š £ P ÿ ™ P ÿ š ÿ † ÿ ™ P š › ÿ † ÿ P ÿ ™ ¨ i .1(c)) (See Fig. 29. 29.1(b)) (See Fig. 29. the process   p  o ¤3QÅ É P p sÔ o p   53o p  o ¤3(¨ É P p QÅ o (See Fig.Æ „ † œ C Æ „ † œ ž š Ÿ ž š ž š Ÿ ž n ³ ¦ p¬3ñé É ² ³ ¦ 3xé ¦ o p ¦o Æ Æ Æ ÿ ž n ³ § p § oñé É ² ³ ¦ 3xé p ¦o Æ ÿ Æ Æ n Æ ¦ ³ Y§ Æ § ¦ ž Æ p § oxéÍp¬3ñé É ¦ o ÿ ª ž šŸ š› « n Æ ³ n Æ Æ ³ § Æ ³ § p § xͬ3ñé o ép ¦o Æ ¦ ÿ p n ¬Ão ’ ÿ ² ¨ ž š Ÿ Æ p § oxéÍp¬3ñé É ¦ o ÿ ª š › « n Æ ¦ Æ n Æ ³ Æ Æ p § ñé o ³ ¦ p n ¨p ¦o ¬1o ’ G3xé ÿ ² § Æ p¬3ñé É ² 1o ’ Gp § ñé ¦ o p n ¨ o ÿ ª P ª š Ÿ ¦ Æ n Æ § Æ Æ ¨ p § ñé o ³ ¦ P ª  p n ¨p ¦o ¬Ão ’ G3xé ÿ ² « ª § Æ n Æ p¬¡xé É ² ¬1o ’ Gp § xé ¦ o p n ¨ o ÿ « P š ¦ Æ ³ n Æ Æ § p n ¬1o ’ ³ n Æ ³ n Æ § ÿ ² ¨ P ³ § « £ ¨ ª Æ p § xé o š p ¦o 3xé ÿ ² « ª p1o ’ n Æ ÿ ² P ³ ¦ Æ p¬¡xé É ² p § ñé ¦ o o ÿ š ¥ £ š « ¦ Æ p¬Ão ’ n p § ñ3xé o ép ¦o ÿ ² P P § § o é p ¦ o ³ n ¬Ão ¦ p n ’ ¨ ›ÿ ² p ñ3xé É ÿ ›ÿ Æ Æ Æ « š † P P P § § o é p ¦ o ¦ n 1o p n ’ ¨ xœ ©ÿ p ñ3xé É ÿ š ÿ Æ Æ Æ « ª † § ¦ p § Èŀ¬3oQÅ o p ¦ p § ñ3xé P P o ép ¦o É ÿ ›ÿ Æ Æ ­ š ©È € † § ¦ p § Q7p3oQ7p § x3xé P P o Å ¦ Å o ép ¦o É ÿ š ÿ Æ Æ  e† € ´ § p § oQÅ7p § oñé P ¦ p3oQ73xé P ° É f Æ ¦ Å p ¦ o £ÿ Æ ÿ š  e† ” € o Ôp …o p 3o … ­ p ‡€V3¤Ô œ © € Q e† œ Remark 29.

For where we have used the fact that is a martingale. Gaussian processes . 291 is ”   ¥£ £ £ £ š   ¥£ £ £   š Æ   p¤3xé É V3QÅ ™ b3sÔ   o p …o p …o ÿ š † † † † ¥ Æ pV3oQÅ7p¤3oxé É …   p …o b3sÔ ÿ ™ š Æ ­ p… V3o p  o ¤3QÅ € © € ™   p  o ¤3xé É p …o b3sÔ ÿ ™ Æ p  o Åp  o 53Q7¤3xé l … i ƒ©jh P š Ÿ Æ p53oQÅ7p¤3xé É °     o ÿ ÿ p… ­ V¡o   p sÔ o œ © ª € . not equal to p ‡Ô o p …o V3¤Ô Å p …o V3¤Ô CHAPTER 29.¥ p… ­ V3o   œ © € ´ b3o p … £ œ neither Markov nor a martingale. The conditional expectation . nor is it a function of alone.

292 .

we get ¬ ¬ g … e† … e† p p o p o u™ p o Ã5ÑÉ ùo Ð ‘ Ï Æ Æ f œ œ ú ³œ p o C œ Ð p o p o ² ÑÉ Ï  Æ ™ Æ  œ œ œ œ f Æ   Ð Ì 7p o ÑÉ  Ï œ ¬ g … Ë ±Æ Then p   ¤3o ú É ÿ P p o ˜ † œ ú p o œ p o ù œ where . so … É ÿ ™ Æ p o  œ p  o ù 53H‚Ð † o ¬ Ï … P p h ° Ð É ÿ ™ 1o  —ÑÉ Ï ¬ ~ p o  … e† œ The process and covariance function From Theorem 1.2) (0. Set Hull and White model Chapter 30 is also Markov. we see that We can solve the stochastic differential equation. and ”p o œ C Consider Æ p o ‘ ú œ p œ p o o ù p o ‘ u™ Æ p o  p o q p H1o  Æ œ œ œ œ † œ œ are nonrandom functions of .69 in Chapter 29. is a Gaussian process with mean function (0.1) 293 p o .f Æ ´     P œ š VŸ p¤3o   ’ ‘ Ð oÏ ’ É † ¬ … p   5¡o ù Ð Æ Ð ÑÉ ÿ Ï ¬ ~ 1Ð š œ  Ï ¬ ‡ ~ p ” 3o … … e† œ i †  ¬ Ï …   P Ð p h É ÿ ™ Ão  ° õÉ Ï ¬ ‡ ~ p o … e† œ „ f ´ ¤3o p   C P Æ Æ C p5¡o ‘ Ð   † oÏ ¬ ” 53o p   P p53o ‘ Ð   † Ï ¬ … É ÿ ™   Æ p   ù ¤3o Ð †  ¬ Ï … P p h É ÿ ™ 1o  p o ÑÉ Ð  Ï † œ ¬ … Integrating.

we define is normal. To do this.3) ÿ ~ p1o ’ ‘ Ð n ’ ž Ï ¬ P … ß ÿ † p¯õo ’ Ô Û €  ø† € œ Æ p o œ  P ß ÿ ~ ® & ­ . Its mean is is (0.n f Æ ’ € ¦ Æ Ð ª Ï ¬ ~ ž P … ß ÿ ~ p1o ’ ‘ Ð n ’ ž oÏ ¬   … ß ’ ÿ ¢ q Æ œ Æ p  o ¤3Hù Ð † o ¬ Ï Ð ‰ ÑÉ Ï ” œ ¬ ~ ÉP ÿ … ß Ð õÉ Ï ¬ P ÿ † ÿ † P p ho ¯Û ” 1r— p h tÛ ” 1o § where Æ ~ … ß œ Æ   ” Ÿ tÛ ” 1j— ¯Û ” 1o 1o p ho p h p h q   q § š í ýü d(ƒë† ž P ’ Q n € ¦ Ð ~ p n ¬1o Ð ’ ‘ Ï ’ ž Æ Æ ªÏ ÿ ÿ ¢ ™ … ß ~ ’ ¬ ¬ … ß P p5¡o Ð   ù Ð Ð p h §1o ‰ õÉ ~ P P É ÿ »q Æ ÑÉ Ï ~ † Ï Ï Æ ÿ ÿ   q I (ƒë† … œ … ß í ýü ¬ ¬ ¬ ß P ’ € p o P ’pd qo ¢ ™ Æ p o  ÿ pd qo œ œ ß € þ í ýü ƒ(ƒë† ¢ œ Æ œ  ß ÿ ~ ® x& ­ ¢ œ Æ p o œ  P ß ÿ q þ ƒü  e† í ý € p h ¯Û ” 1o Ê n f Æ ’ € ¦ Æ Ð ª Ï ¬ ~ ž … ß f ´ p   ù p¯ÛÇo Ô ™ Æ   Æ ¤3o 7Ð oÏ † œ ” p Q1õÉ o ÅÐ œ Ï ¬ P P P ” ´   p¤3oHù Ð Ð   p h p o É ÿ ™ 1o  ° ÑÉ Ï ~ † oÏ Æ Æ Æ ÿ  ÿ œ ¬g… ¬ … ß † xœ œ ß € p o P Æ  „ œ œ ß P p h ° Ð É ÿ ™ §1o  wÑÉ Ï ~ … ~ ¬ ™ ´   Æ P p ÈÅ ÑÉ o Ð Ï f Æ œ œ … ß ÿ † The price at time 0 of a zero-coupon bond paying $1 at time Û p   ù 5¡o Ð ¬ … ¬ †  ¬ Ï … P p h ° Ð É ÿ ™ Ão  —õÉ Ï P ~ P p o Æ  ÿ ÿ … ß † 9œ œ ß ~ p o  … e† œ ¬ p Ûo ¯Ç‡Ô ” ¤3o p   C P p5¡o ‘ Ð   o É ÿ p QÅ † oÏ Æ ¬‹… † œ p o P  ß „ œ Æ œ and its variance is According to Theorem 1.70 in Chapter 29. Then We want to study 294 .

1: Range of values of ”   Consider the price at time … Therefore. Hull and White model Figure 30. 295 … ß ÿ … äá ß † P p ho ¯Û ” Ãr— u = t . In fact.œ  å ’ ” n è Æ € ¦ Æ Because is a Markov process.1) 30. u . this should be random only through a dependence on p o ” Ÿ p¯Û ” oj— ¯Û ” o p o p p ¯Û ” o Ê q   q œ œ § œ š í ýü t(ƒë† œ  ¥£ É £ £ UVp o   p53o   p¯Û ” o £ Æ Ê f  sq ÿ œ þ ýƒõST €e† í ü œ ß ¢ ۔ h ¸ ­ © œ Ÿ ptÛ ” 1oj— ptÛ ” h1o p1o h h p h ¯Û ” 1o Ê f q  q šïíýƒë† ü ” ¦ § Ð ~ P p h ¯Û ” Ão ÿ † § ’ p¬1o n ’ ‘ Ð Ï ’ ¢ q ž ¬ … of the zero-coupon bond: Æ ª Ï ¬ Ð ª Ï ¬ ~ ž … ß ÿ ~ € x¦ Æ Ð ª Ï ¬ ~ ž … ß ÿ ~ p no 1Hù Ð ž o ¬ Ï n f Æ P Ð ~ ž € ¦ p no ¬1Hù Ð p ¤  n wž ¦ 3o ž Ï Æ ªÏ ÿ ÿ ¬ … ß ~ ¬ ‹… ß † † œ ë† † P ‰ ÐÑÉ ~   p¤3oHù Ð   † oÏ Ï Æ Æ ÿ ÿ œ ¬ ¬ … ß ß † Ð   p53oHù Ð   ‰ õÉ ~ P P É ÿ ÿ ß œ Æ Æ †  Ï ¬ Ï ¬ œ CHAPTER 30.1 Fiddling with the formulas T t for the integral. Note that (see Fig 30.

f

p o œ
C ¥£ £ £ £ U V

p o p ¯Û ” o p o p ¯Û ” o Æ bVf ™ Æ Ê  Ê ¬Æ f f œ œ œ † œ
  ¢

p o œ

Æ

’ œ Æ

p ¯Û ” o É
—

œ

q Æ œ

´ ¯Û ” o É p f Æ œ ’œ p o p ” ’ ‘ ¯Û o ’ ¢ q Æ œ œ œ § p p o p o Æ  œ œ œ p ¯Û ” o É p o p o p o tÛ ” o ’ p  Âq  Æ  Æ § œ œ œ œ § œ

œ

œ

œ

§

f Æ

 

p h ¯Û ” 1o p o ˜ p o p ¦ 3o œ § ˜ uq ˜ œ ú É p   5¡o
ž

ÿ

p o p n 1o ˜ «q ˜ † œ
P

”
 

É
„

Æ

p   5¡o

ú
ž

ÿ †

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 Æ Vbf ß ff

n

Æ VVf ß P ff Ð
ª Ï ¬

„

œ

É

f Æ å ’ ” n è Æ
€ x¦

¦

~

… ß
ž o ¬

Ð ÑÉ ÿ Ï

Æ

Ð

ª Ï ¬

~

ž

… ß

ÿ
~

€ x¦

’ p1o n ’ ‘ Ð oÏ ’ ¢ q ž ¬ …

Æ

Ð

ª Ï ¬

~

ž

… ß

ÿ
~

p n 1o ù Ð

… É ¬ e†

Ï

ÿ … äá ß †

p¯Û ” o œ § p ¯Û ” r— o œ
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 ¯Û ” 1j—

p o pœ ¦ ¬3o ˜

˜

Let

30.2 Dynamics of the bond price

to obtain the bond price, its differential must be of the form

p   ¤3o 

É

ß

ÿ q

þ í ý üS € ƒ(ƒ1T e†

p ¯Û ” o œ Ê

Because we have used the risk-neutral pricing formula
—

q Æ œ p o Æ ú œ C p o q œ ’ ¢ q p o œ

É p¯Û ” o p o œ § œ p o p ” ‘ ¯Û o œ œ ù oñp¯Û ” o œ § p ¯Û ” o  Æ œ

 «q q § ° p¯Û ” o Ê q œ † p ¯Û ” o p ¯Û ” o œ Ê † œ

q § „

Ê ¬Æ

” Ÿ tÛ ” or— ¯Û ” o p o p p p tÛ ” o Ê q   q § œ É t(ƒë† š í ýü œ p ¯Û ” o É p tÛ ” o — œ œ
and

we have denote the partial derivatives with respect to . From the formula , we

f Æ

¦

Ð
ª

Ï
¬

~

P

… ß

ÿ †

p h ¯Û ” 1o §

f å ’
ž

p o – ” ­ ۜ Ÿ ¯Û ” p

œ

‘ h © o j—

” n è Æ

€ ¦

Æ

Ð
ª

Ï
¬

~

… ß

ÿ
~

€ ¦

’ p¬Ão n ’ ‘ Ð oÏ ’ ¢ q ž ¬ …

Æ

Ð
ª

Ï
¬

~ …

”p o œ

Ê ¬Æ

p o p ‘ u™ Æ p o  p o œ œ œ œ

ú p o p Hù o ¸½Û œ p h œ ¯Û ” 1o Ê p ¯Û ” o p o p ¯Û ” o Ê q  q É š í ýü tƒ¨† œ ” ¦ Ð œ §~ œ ÐõÉ p ¯Û ” o Æ ªÏ ÿ Ï … ß … œ § ¬ É ¬ ø† ž p n1o ù Ð p ¯Û ” r— o ž oÏ ÿ ÿ † œ ¬ … äá ß ~ ß ”   ¤ 3o ú P p p o É ÿ Æ ˜ † œ ú p o ù1o p o  Æ † œ q œ

f

p o œ
C

Æ

p tÛ ” o œ Ê

Therefore, we must have

We leave the verification of this equation to the homework. After this verification, we have the formula

p ¯Û ” o p o ‘ œ § œ p ¯Û ” o p o p ” p ” ‘ q Æ ¯Û o Ê p o  ¯Û o Æ Ê § œ œ œ œ † œ
.

œ

f

p o œ 

†

p ¯Û ” o É
—

œ

q

’ p o p o ú p o Bx¯Û ” o p tÛ ” o É p o p o ¯Û ” o p p ùo p ’ ‘ ’ ¢ q  «q  q q œ œ œ œ § § œ œ œ § œ

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.

– ” ­ Û h

©

¸

œ

Step 1. From 4 and 5 we solve for

Ê

We take the following input data for the calibration:

1.

– ueuÕh ” ¯Û ” 1o Û i Û i p h



2.

p h Ão

3.

– ÛuiÀÛuijh ” p¯Û ” Ão Ão h p h ‘ § – Ûui ijh ” p o ‘ œ p h §1o ù œ
; ;

5. 4.

(usually assumed to be constant); ; 297

, i.e., the volatility at time zero of bonds of all maturities.

f €¹

¸ k s!ò ”

h h d p k ¤1o p5kh§h h do † ¸ k ˜– s!ò ™Íê h 51o p k
}

€¹

†

t

å ’

n f Æ è

€ ¦

Æ

Ð

ª Ï ¬

~

ž

… ß

ÿ
~

’ p¬Ão n ’ ‘ Ð oÏ ’ ¢ q ž ¬ … q p h ¯Û ” 1o

€ ¦

” Ÿ ¯Û ” 1r— p ho

u

ú ú Û P p¯õo Û   p   ¤3o Æ f ÿ u † ß ” ¯Û ” Ão Û u p h §
u

† q † … e†

p Û ¯õo p Û ¯õo

– ” ¸ ­ Û h eÛ Û ©u ˜
u

Ð 

 

Þ

˜ §

ßÏ
¬

~

p h tÛ ” 1o

¯ ‚ u

Û †u
We can then compute

Step 2. From the formula

§

p h §1o

P Ð ~ ž p no ¬1Hù Ð p ho ¯Û ” 1j— ž oÏ Æ ªÏ ÿ ÿ … ß ~ † ¬ ¬ … äá ß – ” ¸ p ho tÛ ” 1j— ­ Û h eÛ © ¯Û ” 1o p h Ê

  q

š í ýü ¤ƒ¨†
.

We can use this formula to determine

ú Ð Ð ÑÉ Ð ÑÉ p o ÑÉ ’ ‘ Ï ’ q Ï ’ p o p o ù ˆ ™ Ï ’ p f œ † œ œ ¬ … ¬ … œ ¬ … ù ú – ÛuieÛuÕh ” p¯õo Ð ´ p ho ƒ´ ¯Û ” 1r— Û u Ð Ð i Û ’ ‘ ptõo ¯Ço ù ™ Ð Û p Û p Û ¯õo ¡ ù f ˆ u ßÏ ’ q ßÏ ’ ßÏ ’ ßÏ † ¬ … ¬ … ¬ … P ” n p1o ´ ¯Û ” 1j— Û u Ð n p Û tõo ù Ð p ho ’ ‘ Ð oÏ ’ ž Æ ÿ q u ß Ï ¬ ’ ø† ßÏ ¬ … ¬ … ß P ” n Ð ´ ¯Û ” 1j— Û u Ð ~ p1o n p Û ù ¯Ço 1Ð p ho ’ ‘ Ð Ï ’ ž Æ ßÏ ÿ q u ßÏ ßÏ ¬ … … ¬ ¬ … ß ¬ ø† Û ž P ” n °€x¦ Ð U ~ p ¬nÃo p no ù p ho ¯Û ” 1j— u Ð ’ ‘ Ð oÏ ’ q 1HPÐ Ï ž ž Æ Æ ªÏ ÿ ÿ u ßÏ ¬ … … ß ~ … † ¬ ¬ … ¬ 1S ß Û ž P ” n "€ ¦ Ð ~ Ð ~ p¬nÃo U p no p ho ¯Û ” 1j— u ’ ‘ Ð oÏ ’ q Ð ß Ï ~ 1Hù Ð Ï ž ž Æ Æ ªÏ ÿ ~ ßÏ ÿ u … … † ¬ 1S ß – u½ujh ” tõo ù Û i Û i p Û
¬

From assumption 4 and step 1, we know all the coefficients in this equation. From assumption 3, . We can solve the equation numerically to determine the function we have the initial condition .

… ß

¬

…

¬

…

¬

This gives us an ordinary differential equation for , i.e.,

We now have

p Û tõo

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 ù œ

p ho 1Hù

– Èi Û

œ

i jh ” p Hù o œ

œ

o¡ ù Û
u u u u

for all

for all

. Recall that

as follows:

known function of

Ð ßÏ

¬

° °

° … Û Û
u

ßÏ
¬ ¬

… …

Ð
u

…

°

Û

u u

…

ú

P µ ’   p53o   Ð p Û p ¢ õo o ’ ‘ Ð Ï ’ 4 ÿ Ã4 ß Ï ’ ~ ’ ‘ † f Æ  … e† ® x& † ¬ … ¬ ß P ‚4 ”   p53oH7Ð µ ’   ù Ð ~ Ð p h Û ­ ™ Ç4 ß Ï ~ §1o  p ¢ õo  4 † Ï Æ ÿ ßÏ … … †  e†  € ¬ … ¬ ¬ ß p ¢ õo Û 

n f Æ ’ ” n Æ ´
 

€ ¦

Æ …

Ð
ª

Ï
¬ P

~ ÿ Ð
ž 

ž 4

P 4

… ß Ï
¬

ÿ
~

p1o ’ ‘ Ð n ’ ž Ï
¬

… ß
P 4

Æ

p  o ¤3Hù Ð

† o

Ï

~ …

¬ g ž

™ Ð Ï ž
¬  

ß
P 4

Æ

~ 1o ° p h  … p   ¤3o  €

ß

Æ

”
¦

Æ

k Æ

p k ¦ ” Ão f ‰p
t

˜ Âq

 ß „ Û Ì p ¢ õo ”  Ÿ “Û ” ¢ Çr— ³ p’ Ûo ‰ ˜«q q Ÿ D“Û ” ¢ Çr— Í›Û ” p’ Ûo 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 † ” p5kh§h o h d ” €¹ ¸ k s!ò
±h d

 

p ko 5ÃÇ¡
}

q

p ko 511¡
  t

p k 51o ¡ †
  }

ê 

h h d d

i 51o p k
}

q

p k 5Ão
  t

for many values of .

we have

but because

CHAPTER 30. Hull and White model

30.4 Option on a bond

and expiration time

¢ Û

We observed at the beginning of this Chapter (equation (0.3)) that

 

p   53o

4

P

  p53o 4 P   Æ  ß „Ë p“Û ” ¢ Ûõo ¦ ’ ² ~ q § š í ý ¤(ƒü ‚ … ¢ Ûõo p ¢ ÛÇo o zÐ ´ † † oÏ  q § š ý ¤íƒü ´ ‰p p“Û ” ¢ õo o xÐ ’ Û ´ Ê † ˜»q ¢ ut“Û Û ¿ ’

We also observed (equation (0.1)) that

 5~

 5~

‡
† o

 ÿ
4 ³ ² „

~

 ÿ † … € j ø† ~ ° …

t

Ï

‡

4 ³ ² ‚„

€

k

where

is the joint density of

is normal with

ÿ † ÿ † ÿ †

U z 

Æ

p   53o 

P 4

ß

ÿ

µ S ® ·x& ­

¢ †

’ ‘

U z 

Æ

p   53o 

P 4

ß

ÿ

µ ¶ S ƒ

¢


 e† €

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 œ



p o  œ

Û Ì p ¢ Ço



”

 

Æ

p   ¤3o

É 4  ß „Ë

¥£ £ £ £

´p o œ

‰p

˜ «q

Ÿ Í›Û ” ¢ õj— p’ Ûo q

p’ “Û ” ¢ Ço Û §

¢ ” ­ Û h ©

œ

f Æ

¦

k Æ

U ¢

’ ’ ‘ ’
¦

™

’ ‘¢ ‘
¦ 

¢ ™

k

’ ‘ ’ k
S

i

ˆ

¸ ¤’ ¢ p ’i o ’ (i ‘ ‘ xˆ q d ‡ q d ˆ q d d þ ƒü í ý Ÿ “Û ” ¢ Çr— D“Û ” ¢ õo ¦ p’ Ûo p’ Û

‰

³

˜ «q

q

P ”   p ¢ Û ” 3o ‡   i 4 Æ ÿ † ß P   pp ¢ õo Û p ¢ Ço H5¡o Û o pp   p   53o o 4 Æ ­  q   q  ÿ € € © € ß † UVpp ¢ Ûõo p ¢ õo o   ¤3o Û pp   p   5¡o o P ’ ‘ ¢ i ‘  f Æ  q  4 ÿ € S € ß T e†



€

q

The option on the bond has price at time zero of

The price of the option at time



Æ

P  ß „Ë

In fact,

Ì Û 7p ¢ õo

”
 

p   53o

4

where

300

is defined in Equation 0.2.

is jointly normal, and the covariance is

is

(4.2) (4.1)

p ¢ õo Û o Ð ‡ 4 ¹ ´ † † oÏ  q ² „ š í ›ƒ£¥ý £ ü ´ p o ££ ‰ p p’ D“Û œ ˜ «q

~ ° …  e† € ” ¢ õo o Ð ‡ Û ´ Ê
† † o

Ï

4 ¹ ² „

~ ° … €

¸

†

 5~  ÿ

 5~  ÿ

§

q

š í ý d(ƒü

²

~ ‚ …

p ¢ Û ” 3o ‡  

i

¥£ £ £ £ ¢   U e

p o œ

Æ

p   5¡o 

É ß ÿ sq þ í ý üS € ƒƒõƒ e†

p ¯Û ” o œ Ê

¥£ £ £ £ y U V

f Æ p o p ’ œ


 

n Æ q

p n 1o ’ ‘ Ð p ¢ õo o Û  ”
 

p o œ

”
 

Æ

p   ù 5¡o Ð

É
†  ¬

Ï

… ß

4

ÿ

” n Æ

€ ¦

’ å

Æ
¥£ £ £ £

Ð
ª

Ï

èp o ” n Æ ´
 

Æ

p   ¤3o ù Ð

σ

’ ÿ Ð
ž o ¬

†  ¬

Ï

…
ž

Ï

É É Ð P4 ~ ÃÐ ~ 4 ’ ž Ï † oÏ ÿ ßÏ ÿ † ¬ … † ¬ ¬É … ß €ip o ¢   p   53o p ’ o q Æ  4 ÿ œ  S € É ß ~ ƒ e† œ Ð p ¤ 3o ’ ¥ £ ‘ Ð oÏ ’ 4 ÿ ‚4 ß Ï ’ ~ † Æ ¬ … ¬ … e† ß £ ´ p o ££ pp ’ o p ¢ Ço o ° Û p o ’’ ’ q  œ œ   e† œ € Ð Ã4 ~ ™ ÑÉ ‰ ‚4 Ð Ð ~ p o Ï Ï  ßÏ £ ß … … œ † ¬ ¬ ¬ £ ´ p o ££ p ¢ õo ° Û p ’ o   œ € É  e† œ ž ~ p1o n 4 ’ ‘ Ð Ï ’ 4 ÿ ž ÿ … ß ~ † ¬ ¬ … ß É € vp o ¢   p5 ¡o p o ¢ ’ q Æ  4 ÿ ~ á œ  € ß ä É  e† œ ~ ~ p o ° 4 ‰Ð  … œ É … Ð ™ õÉ Ï
U V ¬ ¥£ £ £ ž 

Ï

¬

p o œ

£  

Æ

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

” Ÿ ¯Û ” r— p o œ q

1. Leads to closed-form pricing formulas.

2. Allows calibration to fit initial yield curve exactly.

1. One-factor, so only allows parallel shifts of the yield curve, i.e.,

p ¯Û ” o p o p ¯Û ” o Ê   q § œ š í ýü t(ƒë† œ œ

‘p o ¢ ‘p o œ œ
i

‘



‘


4

ß

ÿ

S ƒ

ÿ † ß p o ¢  e† œ €



can exceed 1.

so bond prices of all maturities are perfectly correlated.

p o œ



301

302

f Æ † … … and covariance function . . for each . The solution to this stochastic differential equation is This solution is a Gaussian process with mean function Define If . 31. Since. is called the Orstein-Uhlenbeck process.1) f ´ 53o p   f p o c ¥C œ c ›C Æ † … ’ p o Æ ‘ ¢ ™ Æ P ’ Ão p h ’ ¢ … É ÿ ‘¢ ™ É p h 1o œ š Ÿ c ÿ ÅÉ ’ ~ œ … ¢ š œ p o … e† œ c ú Å ’ p o ¢ q † œ c Å° É ’ ~ p o … e† œ c „ … ¢ c Å Æ c Å € Å and let be the solution to the stochastic differential equation h ¿ ú p ´ ” h ”  ´ p h 1o ’ Å ™ p h 1o ’ ’ Å ™ 1o ’ ¢ Å p h ff Vbf ™ ” ’ We begin with a -dimensional Brownian motion constants. is a Gaussian process. but (see Fig. The Cox-Ingersoll-Ross model is the simplest one which there is a positive probability that avoids negative interest rates. is normally distributed. Let and h ¿ ‘ p o œ  œ C ff Vbf C ”¢ C o €¹ h l QHp o ¸ p h HÃo œ c  p o Å œ  ” ” Æ VVf d ff Ɔ be º c d † Æ c Å . let be given so that . we have I 'H There are infinitely many values of 303 for which d † Q¼h p o h Q¿  † œ œ Ÿ hQ¿Hp o p o ¢ d ’ Å p o   ´ œ š H€ œ œ † œ ¢ p o ņ ™ Vfbf ™ p o ’ ’ Å ™ p o ’ Å µ p o  f ’ f œ œ œ † œ P »   p ” 3o i … Ð õÉ ‰ ~ Ï … ¢’ ’ ‘ d … † œ and for each . It always has a drift toward the origin.Chapter 31 Cox-Ingersoll-Ross model In the Hull & White model. For .

1) There is at least one value of r(t) = X2 (t) 1 Figure 31. X (t) ) 2 1 x 1 t .¾ ´u t f p   53o ¾ C p   53o ¸ P ¢  p   Æ ¤3o ¾ Å É ÿ p o † œ C ¾ ´u t f p o ¾ œ C p o ¸ ¢  Æ pœ o ¾ Å œ p o œ  Á  ‘ u™ Æ œ Å ¾ ¾ C ¢ ¢ ¾ ¾ C œ » Æ ’ ‘ ™ Æ Æ Æ ¾ C Æ ’ ‘ d » ´u t ™ Ìp o ¾ œ C ’ Æ ‘¢ ™ Æ œ ¢ ¾ ´u t t ú » € p o  q ’ ‘ Æ ~ † œ ¾ ú p o ‘ Æ ´ u t Q™  q œ œ † ¢ ¾ ú ’ ¾ Å ¢ q Ë ¾ Å ˆ ´u t ¢ ¾ ´u t t † p o  Æ † œ Itˆ ’s formula implies o ¾ Å Æ ¾ Å Æ 88‚ ½ ‚½ ’ ¢ ™ ¾ Å Æ 8‚ ½ † h f † h ” k ¾ ˆ 8‚ ‚½ † ½ ˆ þà† ¿$†8‚ t ´t ¢ ’ ’ k ™ VVf ™ ’ k ™ ’ k ff h Q¿ œ p o œ  . 31. Then † ff VVf Ÿ h p o  † œ t š H€ ˆ  Æ If Let 304 . (see Fig.1: x2 can be zero. for which if if ”f º º º † À À p´ k” ” ’ k ” ¢ 1o k Define ( X (t).

We have The Cox-Ingersoll-Ross (CIR) process is given by If happens to be an integer. and they are still correct even when is not an integer.e.. ). is normal with mean zero and variance f p h Ão  Á h † p h §1o p h 1o ´ Å ” h † h Q¿ ~ ®´ ” p h Ão ¢ Å f f bVf p o For example. Cox-Ingersoll-Ross model Then C 305 is a martingale. Take ¢ ‘ With the CIR process. then we have the representation This is not a good parameter choice. f p ‚É ~ … … ú o » p ” o q d ’ ‘ † œ œ i À For . here is the distribution of for fixed . C so is a Brownian motion. {  h † Æ   h Q¿ There is at least one value of for which f † Ÿ h p o  † œ œ œ Æ ” h œ  † ” ” ” p o¾ Å d q Æ ƒbf ˆ d ff œ † p h §1o ’ Å ” h Ão ¢ Å p h † ’ š H€ ù If (i. one can derive formulas under the assumption that integer. then ”p o œ f C p o œ Æ p o œ C  Æ p o Á œ p ‘ Q™ Æ p o  œ œ  Á  ‘ u™ Æ œ ” Æ Æ ú œ € p o Æ q œ ” is a positive ¾ † ֜  ú ù1o p o  Æ † œ  Æ ’¾ Å » Æ ¾ q ’ ‘ Æ  Å  ¢ ¢ ~ ¾ ¾ ´u t ´u t † † Æ C p o  Æ † œ ’ ‘ ¢ C Æ Æ  C š H€ Æ ˆ  C Æ Æ . then h u¿ There are infinitely many values of for which f d † Ÿ h p o  † œ œ l but we do not require to be an integer.CHAPTER 31.. Let be given. If (i. ’ l ’ ‘ ¢ »ù ”p o ¾ Å ’ œ f h Q¿ ’ ‘ ¢ ´u t ù ¾ » ˆ p o  † œ † We define ).e.

2 Kolmogorov forward equation Consider a Markov process governed by the stochastic differential equation f p o œ C Æ p p oÈÇo Å p o Å ‘ ¥™ Æ p Qõo œ œ œ p o We computed the mean and variance of in Section 15. Then ´ p | o {’ z Å y Normal squared and independent of the other term Chi-square with freedom degrees of 31.1 Equilibrium distribution of œ œ p o Thus has a non-central chi-square distribution. The chi-square density with degrees of freedom is is … ¢  ‘ œ  f ’ € p ” o i œ p œo ¾ Å œ ¸ ~ ¢ ´u t ¾ p ” o œ œ i p o  † œ … ¢  ‘ p ” o † œ œ h Ý uÍp o ´ „ t ¥Ý As . The limiting density for p o œ f ’  … f ˆ ¢ ‡ ‘ ~ ª ¢ ‘ { ¢ ¢ ~ Ë ¢  ‘ Å ¤Ä ‘ … ¢ ‘ ¢ … ¢ ‘ ¢ ’ú ³ “‘ Å ¤Ä  ‘ ¢ ¢ Å ¤Ä » ‘ ¦ ² ¢ Ì ¢ ’‘ { Ë Ì ¢ ’‘ { Ë  d Ì ¢ ’‘ Ë { É ¡¢ d É ¢ d ‘ É ˆ œ ‘ { ¢ Ä ˆ ’ ˆ ¦ … ¢   ’ú ³ “‘ ² ‘ ˆ ¢ ’ ˆ d‘ ’ú { m ‘ p ÈÅ o Æ † œ f † p ¦ ¬¡o † »  t † † p o  Ê i ¢ {  ‘ Æ t † uÝ As with . we have . We have x 9 w ¤–Èv 31. œ p h 1o (0.7. and so the limiting distribution of is times a chi-square degrees of freedom. . ‡ ‘ ~ ¢ Ë ¢ p o We make the change of variable .1)  Á p o É ’ ~ … ø† œ ™ … ¢ ’ | € ~ ¢ ›´ Ç Æ { Ç Æ” Å Ä Ã p D¤¤o Ÿ¸u z ¢ ¾ ” y i p o i œ p œ o ¾ Å ~ ¢ D´t ~u œ œ œ ´ „ p o  † œ p ” o i œ œ œ  p o œ ´ œ is normal with mean Å .306 and variance .

Then is random with density variable).2).2: The function . Cox-Ingersoll-Ross model Figure 31. é p ¬¦ ” k ” ” 1o h œ Ê p ho ÃÈÅ œ h p QÅ o  œ h so The Kolmogorov forward equation (KFE) is a partial differential equation in the “forward” variables and .” … Æ ’ ™ P ppV3oQõo p p … o Å #¡¡ ’ … Å p p …o Å p p …o Å ¡ pp ho Åo p o Åo ™ Í1Qõxé p ÈÇxé ’ ‘ b3Qõo é ¢ ™ Æ bgQõo m V3Qõo é É ÿ œ „ € † œ € ” b3o p … pp …o Å pp …o Å V¡ÈÇo ‘ b3Qõo ¡ é É P Æ ÿ C ’ pp …o Å V3Qõo ’ ‘ V3Qõ‚#¡`é ¢’ ™ VŽQõo m ÍV3Qõ‚(é É P ™ Í1Qõxé p Qõxé pp …o Åo ¡ pp …o Å pp …o Åo¡ pp ho Åo p o Åo ÿ „ † œ … Æ ”p o œ C Æ p Íp oQõo p o Å Å ‘ p ÈÇo ¡ é ™ Æ œ œ œ ¦ ’ ’ p p oQõo p o Å Å Å p o Å ¢ p ’ ‘ p Qõo #¡¡ é ¦™ p o õo m Íp Qõo ¡ é „ œ œ œ œ † which vanishes near p o Åo p Qõxé œ Æ †  ¦ Æ p ¦ ” o Ê 3ñé  P p ¦o p o Åo p ÈÇxé ÿ œ † œ € k p QÅ o œ h k p¦ ” o Ê h œp¦ ” k ” ” 1o œ œ We start at at time 0. CHAPTER 31. we assume that (in the rather ¦ Ê p o p QÅ o p ¦o ¬¡xé ¦ µ Let be a smooth function of Fig. Since 0 and will not change during the following. and for all large values of (see 307 œ . Itˆ ’s formula implies o h ¦ h ¦ p ¦o 3ñé ¦  † é for any function . 31. We have  † œ Because we are going to apply the following analysis to the case for all . We derive it below. we omit them and write than .

2.¦ u f at that and nearby h º¤¦ Æ † ’ Uep ’ ‘o Ê ’ u ’ ¢ q p Ê im o ¦ u u ™ É Ê S é P ÿ é f h † ’ ¦ u ’ pp ” p ¦ o p ¦ Ì ¬¦ ” o p¬¡o ’ ‘ Ë u ¢ q ¦ o Ê 3o m o ’ œ œ Ê f h † ¦ Æ U ’ ¦ u ’ pp ” p ¦ Ì7¦ ” o p3o p ¦ ’ ‘ Ë u ¢ q ¦ o Ê 3o m o ’ œ Ê œ ” ¦ Æ ¦ u ’ p ¦o p ¦ Ì ¦ ” o p3o ’ ‘ ’ ¬¡xé  P ¦™ Ë ’ u ÿ ¢ œ Ê | ¦ ¦ p ¦ p ¦o Ì ¦ ” o Ê p3o ’ ‘ ’ u ¬¡xé  P ™ P  Ë ’ u f Æ ÿ œ ¦ ¦ u u Æ ” ¦ Ì p¦ ” o p3o ’ ‘ Ë œ Ê ¦ Æ pp ¦ ” o 3o o p ¦ œ Ê m f Æ ¦ p ¦ ” o p3o ’ ‘ 3o #¡¡ é  P ¢’ ™ ¦ p ¦ ÿ œ Ê ¦ f Æ ™ … … Æ If there were a place where (KFE) did not hold. Integration by parts yields Differentiate with respect to to get 308 of the form in Figure 31. or equivalently. ¦ u u p ¦o 3xé q y † ¦ P Æ Æ p ¦ ” o Ê 3o ’ ‘ ¬3o #¡¡ é  P p ¦ p ¦ ÿ œ p ¦ ” o Ê 3o m 3o ¡ é  P p ¦ p ¦ ÿ œ œ Ê m ¦ † ¦ † œ Æ É p ¦ ” o Ê 3xé  P p ¦o ÿ œ ¦ † Æ p ¦ ” o Ê 3xé  P p ¦o ÿ œ or equivalently. but take to be zero elsewhere. and we would obtain ¦ u u ¦ h j3xé ¿ p ¦o Æ pp ¦ ” o 3o o p ¦ œ Ê m ¦ u u P u ª £ { Dzu ª£ £ p ¦ £)Ì ¦ ” o p3o ’ ‘Ë œ Ê ª£ £ £ u ª £ u | P p¬3o ¡ é  P ¦ ÿ sq  ¦ u u |u P p3oñé  P ¦ ÿ q  u u ª { Dzu y p¬¦ ” o ¬3o p ¦ p ¦ Ê ’ ‘ 3o ¡ é œ † P £ ª£ y { Dzu £ p  ” o 3o 3xé p ¦ p ¦o £ ¤ ¦ Æ p ¬¦ ” o p¬3o p¬3o ¡ é  P ¦ ¦ ÿ œ Ê m ¦ ¦ Æ p … p ¦ ¦ ” 3o Ê 3o ’ ‘ 3o #¡¡ é  P É P ¢’ p ¦ Æ ÿ ÿ p¬¦ ” 3o p¬3o p¬3o ¡ é  P É P ™ Í1Qõxé … ¦ ¦ pp ho Åo ÿ ÿ Ê m This last equation holds for every function ¦ u u Ê S é p ” É ™ ¦ o œ Ê p3oxé  P ¦ ÿ q † ¦ Æ p ¦ ” o É Ê 3xé  P p ¦o ÿ œ Therefore. then we could take points. It implies that (KFE) p ” É ™ ¦ o œ p ¦o ¬3ñé  P ÿ .

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. it will be (EKFE) 309 p Íp o Ê ’ ‘ o ’    ’ ú ’ p ùo ¢ q Íp  o Ê p  q 1o  ‘ † ‘ ˆ q † † † p Ç#¡¡ o  Ê ú ’ ‘ ™ q ˆ o  ’!‘ ™ p o  ˆ ” ú ’ p ù 1o   q ’ ‘¢ q d ú ’ ù Ì Ê  q ’ ‘ ¢ q Ë ú ’ p o Ê Ì   q ’ ‘ ¢ q p o ’!‘ ú  q p o Êf ’   Ê ˆ ’ q ² ˆ ’  “‘ ’ ˆ ù ’ Ë ˆ ’ d‘ ‘ q ù p Ç¡ o  Ê † À§ . we must have CHAPTER 31.f h † u u u u p á o  Ê ú ’ p o ³ p ù 1o  Ê ’  q ’ ‘¢ q ú ú ’ ù  ’“‘ ™ p o Ê p Ì  q ’ ‘ ¢ q Ë  q ˆ We want to verify the equilibrium Kolmogorov forward equation for the CIR process: f Ì ¢ ’‘ { Ë É d ¢ ¢ Ä ‘ ’ú ³ d‘ ² ˆ ” ‡ ~ ¢ ‘ Ë ¢ … ¢ ‘ ¢ ‘ ¢ Å ¤Ä  § ¨† Æ f d † ” h ¦ ™ò ¦ p ¦ 3o Ê  ÿ  p ¦ 3o Ê P  h f h † ¦ u u ¦ u p ¦ p ¦ pp ¦ p ¦ Ì ¬¡o Ê 3o ’ ‘ ’ ¢’ ͬ3o Ê ¬¡o m o Ë ’ u q f p ¬¦ ” o É œ Ê  ÝÉ •– †Þ † p ¦ 3o † h f p ¬¦ ” o œ Ê  ÝÉ •– †Þ Ê œ If the process p QÅ o We compute p o  Ê œ Letting t QÝ where 31. we obtain the equilibrium Kolmogorov forward equation has an equilibrium density. Cox-Ingersoll-Ross model We computed this to be in (KFE).

The bond price process is is given by . p ¯Û ” o œ  p ¯Û ” ” o ¥£ £ £ £ ¢   œ  p¯Û ” ” p o o p tÛ ” o Ê f  Ê œ œ † ” œ p ¯Û ” ” o p¯Û o  Ê p œo Ê  Ê œ œ p   53o  ß ¥£ £ £ £ ¢   P ۔ ” œ  ” Up o œ Æ U V ÿ q p ¯Û ” o œ Ê Q   f p o œ Æ p   53o  É ß ÿ sq þ ƒõT e† í ý üS € ú p ¯Û ” o œ Ê ”p o œ C Æ p o œ   Á R p ‘ Q™ Æ p o  œ œ q ù1o p o  Æ † œ œ  ´ p ú ’  q ´ p ú ’ ’  q ’ ‘ ¢ p  p o ¡¡  ú ’ ’ ’ ù ¾‘ p ’ ‘ ¢ q 1o  q  q ’ ‘¢ q ˆ ú ’ p ù ’ 1o  “‘  q ’ ‘ ¢ q ú ú ˆ ’ p ù ’ 1o  !‘ p  q ’ ‘¢ q  q ’ ú úˆ ’ ’ ù1o  “‘ ™ p q q  q ’ ‘¢ q ˆ ú ú ’ ù ’ 1o  d‘ p ’ ‘ q ’ ‘ ¢ q q t q ’ p o ¡ˆ p o¡ p ’ ‘ ¢ q  ʒ ‘ q  Ê  Ê p Ç#¡¡ o o ‘ p Ç¡ f  Ê  ’ u™  Ê ’ ‘ ˆ p o p Ç¡ Ê ’ Q™ p o Ê ’ ‘ o  u † Íp o Ê ’ ‘ o ’  u p   ‘    u ’ u † ú ú ” p o¡ p ú p Íp o Ê p ùo 17o    q  Ê 31. i.4 Bond prices in the CIR model ” h † ù 1o  ™ d ’ ’ ‘¢ q ’ ù ‘ ¢ q 1o ° p  o Ê †   o o Ê Ê ù 1o  ™ d ú ù1o ™ ° 5p q ú ù 1o ™ p q ú † q  q ù1o ™ p o  Ê q † as expected. The LHS of (EKFE) becomes Now 310 The interest rate process p o u u where p h 1o Because þ ƒü S í ý  e† € Æ p5¡o É P    ÿ q I í ý ƒü is given.. The Markov property implies that is random only through a dependence on .۔ ”p o œ œ Ê the tower property implies that this is a martingale.e. Thus. there is a function of the three dummy variables such that the process is the function evaluated at .

we look for a solution of the form ” ul ijh h Û f  ‡ ” hœ tÛ ” ” o P‡ ’ ¯Û ” ” o ‡ p ú  p p ù 1o p ” ” É p ” ” ™ tÛ  o Ê ™ tÛ  o Ê «q Ê ’ ‘ ¢ ™  Ê  q  œ œ œ † œ  Æ Setting the term to zero. its differential has no term. we obtain the partial differential equation C Æ   ‘ œ É ‡ 7‡ ’ ¢ ™ ™ ‘ ‡ f œ Æ Ê p œ úÆ  ’ 1o ‡ Ê ù Ê ™ Æ  q Ê ™ ÀÊ  q Æ  œ œ † equals ´ ¯Û ” ” p o o É p p o f Æ  Ê ™  œ ‡ œ œ œ ” ” ” ” ™ p o  Æ p¯Û p o  o Ê ™ Æ ptÛ p o  o œ œ œ œ œ œ ³ tÛ ” p œ p o ¯Û ” p Æ  Æ œ œ p o ° Q   Ê  q  œ ” p o o R  Q œ  Ê ‡ ” p o o 7‡  œ p53o É P    Æ Æ p¤3o É P    ÿ q œ Æ œ œ  Ê Æ  „ q d í ý (ƒü Because pute p ¯Û ” ” p o o f   The terminal condition is f (4. Using the Hull & White model as a guide. Then we have is a martingale. this equation has a closed form solution.1) h   p tÛ ” Û ” o  Ê ­ bVf f f ’ © Ê ¢ ÿ sq I í ýü (ƒë† í ý (ƒü ² Æ I ) p   ¤3o É P where h p ¯Û ” õj— ” h Ûo p ¯Û ” õo Û and then set We first solve the ordinary differential equation and the partial differential equation becomes The expression in CHAPTER 31. We com311 .” h † p tÛ ” Ço Û § É ”   p¯Û ” 3o   ù ¯Û ” j— p o Æ ÿ § ß † œ ú ’ p ” ž`h p ” É ™ tÛ o ’ ‘ ¢ ™ ¯Û o q d q œ § œ § † p tÛ ” o ’ œ § ʒ § ú É É ’ p ù ™ 3o — p o ¢ ™ Ê ú q ’ ’ ‘ ¦™ q d q Ê  §’ † h p ù Ão § p É —§ É § o ™ Ê  q ’ ‘ ¢ ™ Ê  q q Ê q   q  § § † ‡ 7‡ ‡ ” ” ʒ Ê Ê q Ê § ” ë† p É — É § o † É Ê Ê q   q § † ” Ð ¡É 2 ß Ï ˜ ~ 2 ‡ ÃÐ ¡É (~ p ¯Û ” ” o  Ê ß Ï } e† … œ ” d † † † § Surprisingly. Cox-Ingersoll-Ross model .

Û   h ¿ p   Qw5¡o . and we have p ¯Û ” j— o p ¯Û ” ” œ o  Ê œ œ § and It is tedious but straightforward to check that the solutions are given by so 312 and and are given by the formulas above. Moreover. and . Similarly. Because the coefficients in depends on and are functions of only through their difference . the function . so is strictly decreasing in . we have ˆ f ~ ™ † † † … … åp Û ” è p q Ço o œ … ” œ œ o o j— § h † p tÛ ” Çr— Ûo ú ’ ” ”   ±   ’ “‘ ˆ 9’ Á ¢ ™ ~ ˆ † Ì ê –§ Ì ˜ – q † † †ê … … ú ’ p Û ± ¢£™ p q õo o ’ “‘ p ¯Û ” ù q ́†Ñ(~ ˜ – Ðê É … $ … á Ì ê ’ œ   † ßÏ … ¢ ä xoÞ ˆ … “… ú ’ p p Û õo o p Û Ço o ± ¢ ™ Íp p tÛ ” q q p Û œ … ˜™p Íê õo o œ Ì– … – … Ì ê q † œ … ™Íê Ì ˜ – p ¯Û ” o É — f p ¯Û ” o œ § ù q † œ where † ¤… where   ” ul Û œ i jh ” Ÿ ¯Û ” j— p o œ q p ¯Û ” o œ §   q š í ýü t(ƒë† do not depend on . We write instead of p ”” o  Ê p¯Û ” ” o  Ê œ q Û œ p ¯Û ” o where ú ’ ’ ™ f “‘ ˆ 9’ Á ¢ † ¤… ± ™ p ” o ’ “‘ p ” j— o ù q … – … á Ì ê   † ä xoÞ ˆ ú …“… ’ ± o ¢ £™ p ” o p ”o … ́†–ê p ˜ … – … Ì ê ” o † § … †ê Ì ˜ – We have ß ÿ q þ í ý € ƒƒü  e† p p h ¯Û ” §1o o  Ê Now for each . almost surely.” p ”p h d §h Ão  o Ê † p p h ¯Û ” 1o o  Ê P  f ¢   Æ p   5¡o  å ” èp ” o … Ì  ú ’ ¢ ˜– †ê ˆ ’ … ¢ ” p ” † ” ” h ” h ” Ÿp p ”” o ” ” or— q p ” o  q    Ê   § t(ƒë† š í ýü Û p tÛ ” o p ¯Û ” j— o ” q Û p ¯Û ” ” o œ † œ §  œÊ œ œ ú œ p o p o p p o ù1o p o Æ  Á‘u™ Æ  q  Æ † œ œ C œ œ œ ” h ¥£ £ £ £ ¢   É ß ” ¯Û ” ” p o o p œ œ  Ê U 3 † p o œ Æ p   53o  ÿ q þ ƒõT € í ý üS Thus in the CIR model.

and . so and is strictly inreasing in .6 Deterministic time change of CIR model p o Process time scale: In this time scale. is a martingale. 31. the interest rate is given by a time-dependent CIR equation i ¢ ui Û œ i Õh ” Vp o U ” ‰p h p’ ۔¢ ۔ o o p ” ¢ Ç!n Ûo f  ˜ Âq  Ê   † p ” !n o n  ú ‡ 7‡ œ † ” ¢ Ûul ijh ” h ’ h n ’ ‘ ¢¦™ ‡ p n ù 1o ™ É n ™ n f    q   q  œ † p p o ” d   53o É P o n f p   ¢ Û œ ¥£ £ £ £ f h ú p p oÍ p o Í p Í p o zp o p HÍ ù o o pÍ o Í Í o Í Æ Í  Á Í ‘ ™Í Æ Í  Í Í q Í f  Æ œ C œ œ œ œ œ œ † œ pÍ o Í  œ ú p o p o p ù p o ‘ Q™ Æ p o  q 1o  Æ œ œ † œ ‰p † ” Ÿ p¯Çor— p¯Ûõo p1o Û h p p h ¯Û ” §1o o q  q  Ê § ští(ýƒüë†  Ý ß p p h ¯Û ” 1o o f ˜ «q Q   œ p’ “Û ” ¢ õo Û Æ  Ý ” t ß o p Û p h ­ p¯ÛÇr— ™ ¯õo 1o ” † h pÃor— p h p h  •– h ™ 1o § 1o©  †Þ § C p¤3o  P    ÿ sq Æ œ  Ê œ Á R ptõor— ™ ¯Ço Û p Û o †  ¢   I Æ í ýü € (ƒ« ø† p   5¡o ’ “Û  § 4 p h Ão ß É œ ÿ q   œ þ í ý üS € ƒ(ƒÑƒ e†  Æ Ê •– †Þ  p p o ” dn o œ Û „  q d · œ œ í ý ƒü ’ “Û . As usual.) 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. the interest rate equation is given by the constant coefficient CIR Real time scale: In this time scale. Cox-Ingersoll-Ross model 313 But also. is the maturity time of the bond. 31. and this leads to the partial differential equation (where .CHAPTER 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.

. . . . . . We have and this will be if we set ” ¢ Í œ Æ Í p Í Ñ¡ p Í o o o p œ pÍ o ¡ Ï œ f Ï p Í Ç¡ o œ  Ó ß Ï pÍ o P p p Í o o ÿ q œ Ï þ í ýü € ƒ(ƒ« ø†  pÍ o Í  † œ p tÛ ” h ” 1o o p h  Ê pÍ —Û ” h ” 1o ¤o p h Í  Ê p  Û o Í Û With . . ” Ð 3É 2 ß Ï Í Û denote the price at real time of a bond with maturity Let is . . and and are as defined previously. . We need to determine the relationship between and . . . . Í Î " which relates the two time scales (See when the interest rate at time Í œ œ µ .3). . . . . . . 31. . . . . . . . . . . . . . . We want to set things up so ˜   p ¯Û ” j— o p ¯Û ” o pÍ tÛ o œ œ § ~ 2 ‡ 1Ð 3É (~ p ¯Û ” ” o p  ” ”Í Í Û Í ¤o Í  Ê  Ê ß Ï } e† … œ † œ pÍ o There is a strictly increasing time change function Fig. Ï Ñ p o Í . .314 A period of high interest rate volatility Figure 31.3: Time change function. . . . . . . . . . . . . . . . œ † œ Ï l † 9œ œ µ Ð Ï l † Î " œ ¶ œ pÍ  Û ” Í ” ¤o Í  Í  Ê Ï Ò Ï Ò Process time Real time œ † ֜ † Í Í œ in the first integral to get . . . . . . make the change of variable p oÍ P pÍ —Û ” h ” 1o ¤o p h Í Í Æ Í  Ôsq f Ó ÿ  Ê œ ß þ ƒ» ë† í ýü € ¢ œ P p o p tÛ ” h ” 1o o p h  Ê . . . ” œ ¢ Æ œ Æ œ œ Ï Ò  † ֜ ß ÿ q Æ œ þ ƒ» ë† í ýü € Ï Ò Í † ֜ Û ”p o Í where . . .

.Ï † Ï Ï ” p7’!Í Û ” h ” ph1o o Í l Hp ¢ Í Û ” h ” 1o o Í p h  Ê q  Ê q   xoÞ   xoÞ ’ “ ¢ h 1o p h Í Û l Í Û ” ¥H—Û ” h ” 1o ¤o t l pÍ p h Í i Õh Í Û  ÍÊ q †  Þ   Ý ß t p ho ¯Û ” 1j— p ” h p h ™ ¯Û Ão Ão Í  ­ f •– p ho h ” Ãr— ™ h ” ço é o Í †ØÞ p h p h  ”† h † § Ï f Í p Íp Û o ” 1oj— ™ —Û o ” 1o h ppÍ h Í Ï Ï Û ‘§ f p o  œ pÍp—Û o ” §h1o oj— ™ —Û o ” 1o o p ppÍ p h Í Ï Ï Ï Ï § p1o ¡ h p h 1o Í  Ï p Í Û ” h ” Ão ¤o p h Í †  Í Ê    Þ q or equivalently.7 Calibration CHAPTER 31. Cox-Ingersoll-Ross model Suppose we know and only through for all ¸ Í eÛ Take and the functions equation for The right-hand side of this equation is increasing in the having limit at . .e. this solution 315 and (*) . Take (we justify this in the next section). in the real time scale. These values determine . We calibrate by writing the equation variable. then is a strictly increasing time-change-function with the right properties. Í Û Í œ where 31. If . For each . so the equilibrium distribution of seems reasonable. Thus . ” fgp ” h p Û ” h1o h pÍ —Û ” h ” 1o ¤o Í p h Í Í Û 1or— Í Í p§1o Í q Í  q  Ê § d·íýƒüë† p Í Û ” h ” 1o ¤o Í p h Í p h §1o Í   Ê – ” ­ ÍÛ h © Û Í q Í œ ppÍ HÛ o ” p Í o j—  Û o pÍ † p ÍÏ o ¡ œ Ï pÍ  Û p p Í Û o œ ” p Í Ï o o † Ï ” oÍ Í j— œ ”Í o Í œ § œ Ï § ppÍ HÛ o ” p Í o j— o ¢ Ï œ Ï ” fgp Í Û ” Í j—  Û oÍ p q Í œ pÍ o ¡ p q p Í Û o œ ” p Í Ïo Ï €i Í p  Û o œ Í Ï ”Í o Í pÍ o Í  q œ § œ d í ý ü e(ƒë† p oÍ o Í  q œ þ í ýü ƒ(ƒë† Ï § ¡ ” p o ” pÍ o p  ” ” Í Í Û Í p Í o ¤o Í pœ o Í Ï ~ Ê  Ê  † œ œ œ Ï Í œ do not depend on and are time dependent. the model coefficients . i. since. solve the : . starting at 0 at time h pÍ  Û o t t p h 1o Í §  † p —Û ” h ” Ão o p h Í Í¤ Í  Ê    Þ q d p h Ão ¡ pÍ ¯Û o ” — Ï † Ï ú” ù Since is (*) has a unique solution for each . For h Í Û Ï so p’ “Û o Í l  p ¢ Û o Í .

Í Û pÍ —Û ” h ” 1o ¤o p h Í  Í Ê   xo Þ q t ÝÍ Q¯Û ” h  Í Û ”  Û o j— HÛ o o ppÍ o ™ ppÍ Ï Ï p h 1o Í p h 1o ¡  d Ï § † pÍ —Û ” h ” 1o ¤o Í p h Í  Ê f p h 1o Í P Ó £ u £  † ß ££ —Û ” h ” 1o o p p h Í Í¤ Í Í Û u  Ê    Þ u Ø q In the real time scale associated with the calibration of CIR by time change. we write the bond price as ” —Û ” h ” Ão ¤o Í pÍ p h Í  Ê £ £ p h 1o  P † † † u ß £¤tÛ ” 1o £ p h Ê Ê Ê Ê u u Û u u ´ †   ¢ p h ¯Û ” 1o p h ¯Û ” 1o Û   xo   xo u u Þ q ´ z P Ð ‡ † o f ´xÐ ‡ ~ ³ † oÏ ² „ … j € p Û ~ ¯õo ° Ï ³ ² „ Æ f p   5¡o    ¢ …  € q † Þ q ß Æ ÿ q  ß p h §1o f  þ ƒü  HxoÞ í ý €   P   xo Þ p   53o ÿ q P þ í ý € ƒ(ƒü  e† ß ££ ¯Û ” 1o p h u £ £ p h ¯Û ” 1o Û Ê    † Þ q 31. p h 1o Ï .4. starts at 1 when and goes to zero as . is strictly increasing in . Therefore.8 Tracking down x × w Ö $G)Õ Consider the function § p h 1o Í  § pÍ —Û ” h ” §1o ¤o Í p h Í  Ê Þ q h   xo The calibration of CIR by time change requires that we find a strictly increasing function such that † q Here thereby indicating explicitly the initial interest rate. 31.ú ’ f ’ ‘ ˆ ™ ’ Á ¢ † ‡… ú ’ å ± ¢ p ¦™ ¯Û o ” p¯Û o ’ ‘ p Ûo ¯õj— è ù q … ́†ê ˜– … $ … á Ì ê ’   † ß … Y… ¢ “… ä xoÞ ˆ ú ’ p p ± ¢ ™ tÛ o ” ¯Û o p Û ¯õo … †ê ¯Û o Ì ˜ – p … – … Ì ê … Ì  ” ¯Çr— ¯õo p Ûo ™ p Û h † Û Í Ï where . The above says that Justification: 316 Result for general term structure models: and are given by ˜– †ê † § p Ûo ¯Çr— p Û ¯Ço in the time change of the CIR model (cal) with . is as shown in Fig. determined by market data.

. .... . As . . Cox-Ingersoll-Ross model 317 Strictly increasing Figure 31. . . p¯õoj— ™ tõo Û p Û t § p h §1o Í Û Í  µ Ï § p h §1o h Í  † † Û p —Û ” h ” 1o o p h pÍ —Û ” h ” 1o ¤o p h Í Í p ÍÛo ͤ Í   Ï ptõor— ™ ¯õo Û p Û Í Ê Ê   xo   xo Þ Þ q q ¶ ¶ § p h 1o Ï Goes to p —Û ” h ” 1o o p h Í Í  ͤ Í  Ê Í t Ý u¯Û   xo Û Þ q . . This is because the interest rate is positive in the CIR model (see last paragraph of Section 31.. If defined a time-change function than (cal).5) t Û f pp Û o oj— ™ p Í Û o o p Í Ï Û µ ...4: Bond price in CIR model Figure 31. let us first consider the related equation (cal’) t Ýp u¯Û o Í t Ý Q¯Û Í p’ Ûo Í l ‘p ¢ Û o Í ’ Û lÏ¢ Û Í If . . . .. . ... except it satisfies (cal’) rather Ï p Ûo ¯Çr— ™ ¯õo p Û § p h 1o Ï Í  † Ï pÍ —Û ” h ” Ão ¤o p h Í Û  Í Ê   xo Þ q h p Ûo † Í p  Û o Í Ï h † Í Û Fix and define to be the unique for which (see Fig. . . .. .... Í . . To solve (cal). .5: Calibration The function is zero at . then .. .. .. . .. . ...4). . then . . 31. ... . ... . . . ..... is strictly increasing in . . . ... ... and goes to as .CHAPTER 31. ...... . . .. We have thus which has all the right properties...

observed in the market. Dividing by .h f † ’“‘ ù q ˆ † ’“‘ p ho 1‚¡ ù q † ˆ — ± Ì 7p ” o … Ì  ˜– †ê ú ’ ¢ ™ p ” o … $ Ì ê … ˆ Ë ’ ˆ ´p ú ’ p ¢ ™ h1o q h ™ o … … … ú’ ” Ì 5´ 7p o ± ” ¢ £™ p ” … $ê … Ì Ìp ” o … †ê Ì ˜ – … … ˆú° ’ … U — p ” o … Ì  ú ’ ’ U ’ ¢ q ˆ ú d… † ’ S … p h ™ o … ´ ° ˆ úÍ° ’ d … h …™ … … ˆ o ’ Ë ’ … q ˜ … ›… ú … ́™–Íê … ˆ ± ’ ¢ ™ p ” o Ë … $ … ö Ì ê d ’ú ˆ ’ … ± ¢ … ¦™ “p … ” o S ˆ ˜– †ê … Ì –  ê  ’“‘ p ”o¡ ù q ˆ † — ´ Ìp ” o … $ Ì ê ± ± Ì –  Ìp ” o … Ì  ˜– †ê ú ’ ¢ ™ p ” o … ê § … Ë p ” o … ’ ú’ ’ p ho p p ho 11¡ ¢ £™ 15h q h ™ o f d … † … … „ d † … § ú ’ ¢£™ p ” o ’ Ë p ” o q … … ™Íê … ú … †ê Ì ˜ – Ì ˜ – ’ ± Ìp ” o ¢ ¦™ p ” o Ë ± ° ’ … Ì  ˜– †ê Ì $  ê  … d … Ì –  ê  … f d p h 1o ¡ † Ï § — p h §1o ¡ d † § p h 1o Í  § 1o Í † p h ͧ Û p h p h 1o ¡ Ão Í p h Ão Í f   Ï † h 1o ¡ p h † p h p ho¡ 1o1¡ 11 — ™ 1Ñ¡ ÃÇ¡ p ho p ho f ph1o ¡ ph1o o Ï ¡ — p h p p p ™ 1o ¡ 1Ïh o o ¡ £ Ó £ Ï Ï Ï P Ï ß£ p —Û ” h ” §1o o p h so also satisfies (cal). so we have u £¤ Í Í¤ Í u u Ø   Ê   xo Þ q † Ï d † Ï We conclude by showing that † p h §1o ¡ p h 1o Í  Computation of p h 1o ¡ Computation of … p h 1o ¡ Note that is the initial interest rate. and is striclty positive. † p ” ‚¡ o § p h 1o Í p h 1o Í   . we obtain We show in a moment that 318 : : . From (cal’) we compute .

differential equations be given. and let Yield at time on a bond maturing at time and p YªÅ o ’ p o¢ Å h ‘§1ªÅ ½ƒÃo ¢ Å ¿ p ho ’ h ¿ p h where and ’ ¢ o ’ Å Æ œ ¢ o Å Æ and Then is a Brownian motion with are independent Brownian motions. To simplify notation. we define 319 ’ ªÅ Æ ’ Å Æ ” œ Æ Ô ¢’ ‘ † ¢ Å Æ ¢ Å Æ be given by the coupled stochastic (SDE2) (SDE1) .f Æ œ Ô ’ ‘ ¢ (i ‘ † ’ ªÅ Æ ¢ Å Æ ” œ ” œ Æ i Æ Ô ’’ ‘ † p o † œ Ù C Æ p o¢ œ C Æ Ù C p o ’ ’ i q d Á ™ p o¢ i µ p o f ú œ C ú † œ o ÚC Ù ”œù C ™ p o ’ Å ’ ™ p o ¢ Å ¢ µ p ‡Ô œ œ † œ C C ” p o ’ p œ ú Á Á C Æ ’ i q d Á ú ™ p o ¢ Æ i o»ù ™ p o ’ Å ’ ™ p ú œ” ¢C œ p o ù ™ p o ’ Œ ™ p œ œ C Æ o¢ Å¢ œ ¢ ¢ ú o Å œ k k ’ ‘ u™ Æ p ’ œ ¢ ‘ u™ Æ p ¢ œ m 5 p o ’ Š͒ ’ m5™ œ ™ p o ’ Œ¢ œ p ™ p o ¢ Å¢’ o œ ¢ k † p ™ p o ¢ Å ¢ o œ k † œ œ œ ” u™ P œ A two-factor model (Duffie & Kan) Chapter 32 Let us define: Interest rate at time œ œ p o ¢ Å † œ p o ’ Å † œ Let .

2. (SDE2’) (SDE1’) p o f ” p œ o ¢Ù C C p o Ô Æ œ p o Ô Æ œ k k œ ’ p ‘ u™ Æ D’ œ ¢ ‘ u™ Æ p ¢ œ m 5 p ªÍ’ o ’ Œ m5™ œ o ’ Å ™ p ªÍ’ ¢ œ ” ¥l t œ i ±h p ™ p o ¢ Å¢’ o œ ¢ k † p ™ p o ¢ Å ¢ o œ k † œ ” QHp sÔ h ¿ o œ ú o ’ Å Æ œ ¢ o Å Æ . We shall choose the parameters so that: ’ ’ ú ¢ ‘ ú ’ ‘ ’ ™ ’ ‘ ú (i ’ ’ (i d ¤’ ™ p ‘’ q ¸ œ ¢ ú ™ ¢ ‘¢ ú Á ˆ ’ ’ ú ú p ‘ ƒ’ i ’ ™ ¢ ‘ ¢ o o¢ C œ ™ ’ ‘ ¢ (i ’ ‘ ú ™ ’ ¢ú ¢ o k ‘i ’ ™ ¢ ú p¢ m¢ Æ ú ¢ ú ™ ¢ ‘¢ ˆ ’ ’ Å ™ ¢ 7p ¢ ú ’ ’ ¢ ¢ sÔ ok Æ ‘ ú C ™ ’ªÅD’ ¢ ¢ ’ ªk Å ’ Æ ú o ™ ú ¢ ™ ¢ ¢ k  ™ ¢ Å ¢ ¢ ¢ ú™ ™ ¢ Å kÆ ú o © † ú o ¢ ú † Ô † Æ 32.1 Non-negativity of Û where 320 Assumption 1: For some . These can be rewritten as ”¢ ú From our discussion of the CIR process.Á Á p’ ’ f ’ ‘’ ú ú ú ¢ ¢ ú ’ ™ ’ ‘ ¢ i ’ ¢ ˆ ™ ’ ‘ ’ o ¢ ‘  ” hQ¿»ù Ão ªÍ’ ú 1o ¢ Å ¢ ú ™ p h ’ Å ™ p h Ô † will stay strictly positive provided that: f  C Æ Ô  ú ú ú ¢ ¢ ú ’ ¢ ’ p’ ’ ‘ ’ ¢ ˆ ™ ’ ‘’ o ¢ ’ ‘ ’ ™ ‘ i ú Ô p’ ’  ’ ’ ‘ ’ Æ Â ¢ ú C p ù ’ ’ ™ ¢ m œ Æ … q m f … ë† ú ú ’’ ’ ™ ’ ¢ ¢ k k … ¨† ¢ ’ ’ ú ™ ¢ ¢ ú ¢ k k p ’ o œ C œ Æ p D’ m ’ p o Æ ú œ  Cú p ’ ’ ™ ¢ m¢ o Æ m œ D’ p ’ ‘’ Æ (i q ú C ú ’ Œ ™ ªD’ ’ k ú p ‡Ô o p’ ’ Á ’ ’ ‘ ’ ¢ ú œ ’ Åp’ ™ Æ ­ ú ª7D’ ’ œ k d Á ’ ™ ¢ Æ ’ úC ™ ¢ Å¢’ ’ o ™ k is a Brownian motion. almost surely. and 3. ’ ú and (SDE1) and (SDE2) make sense. we recall that ú p ù ’ ’ ™ ¢ m¢ ™ Æ q m ú ú œ … ¢ ¢ ™ ’ ‘ ¢ (i ’ ¢ ˆ ™ ’ ‘ ’ ‘ ú ¢ úo ’ Å ™ Æ ­ ù ™ ª7’ œ … … … p o † œ  C Then Assumption 2: p ho §1‡Ô ú o ™ Æ Ô ú œ … ë† o ™ ú Ô ™ ¢ Å¢ Æ #…© † and … q ™ ¢ m¢ Assumption 3: ù ’m’ ú Under Assumptions 1.

2 Zero-coupon bond prices CHAPTER 32. this is random only through a dependence on . Thus. We compute its stochastic differential and set the The usual tower property argument shows that 32. the bond price . Since the coefficients in (SDE1) and (SDE2) do not depend on time. there is a function depends on and only through their difference of the dummy variables and . 321 (*) f h ¿ ¥Âù ™ ¤ƒ’ ’ k ú ™ ¢ k¢ ’ ¤k ” ¢ k h í ý ƒü . A two-factor model (Duffie & Kan) and all of a zero-coupon bond paying $1 at time satisfying term equal to zero.ú ” Ÿp o o ’ p ” ” ¤k ” ¢ Ão ’ k ” r— q p ” ’ 5k q p ” o ¢ ¢ k q Ê § § š í ýü t(ƒë† (PDE)  ” 4 ‚ ‚ Ê 4 h f p xù ™ ˆ Ê q ‚ Ê ¢ p ’ m s ™ ú ú ú ú 4 ’ pxù ’ k ’ p ¢ ¢ ™ ¢ k ¢ o ’’ ‘ ¢ ™ ¢ ‚ ‚ Ê xù ™ ’ k ’ ™ ¢ k ¢ o ’ ‘ ¢ (‰™ ‘ i ú‚ ‚ Ê ú ™ ¢ ’ 4 ˆ ¢ ’ 5† kƒ’ p ’ k’ ’ k ™ ¢ k ¢ o ’ ‘ »™ ¢ ‚ Ê ƒ’ '™ ¤ƒ’ ™ ¢ k ¢ ’ o ™ ‚ Ê p ¢ '™ 5ƒ’ ¢ ™ ¢ k ¢ o ™ Ê q Ê ¢ k q ¢ m m k k k k p ” ” 5k ” ¢ 1o ’ k Ê 4 ´ Ô ’ ‘ ¢ Ô ¢ ‘ ¢ Æ ‚ Ê  u™ Æ ‚ Ê  u™ Ù C C 4 ³¢ ¢ Ԓ ’ Ô ’ ‘ ¢ ‰™ 4 4 Ô¢ ’ ¢ ‘ i Æ ‚ ‚ Ê ’ ‘ ¢ ™ ‚ ‚ Ê ‚ ‚ Ê ’ ‘ ¢ ™ œ 4 P ¢ ° ’ªÅD͒ ’ Å p   ™ ¢ Å ¢ ’ o ™ ‚ Ê p ¢ 5™ ’ ’ ¢ ™ ¢ Å ¢ o ™ Ê ¢ Å q ² «Q   Æ ¤3o ¢ Å É sq m ÿ I (ƒë† k k k k í ýü ´ ’ ’ ¢ Å ¢ Å 4 4 ¦™ ’ªÅ ’ªÅ ¢ ¢ ¢¦™ ’ Å ¢ Å ¢ 4 Æ Æ ‚ ‚ Ê Æ Æ ‚ ‚ Ê ™ Æ Æ ‚ ‚ Ê ¢ ˆ ¢ Å 4 ¢ Å «Q   ¤3o ¢ Å P ° ’ªÅ ¢ p   É sq Æ Ê q Æ ‚ Ê ™ Æ ‚ Ê ™ ½Ê Æ Æ q ÿ I (ƒë† œ œ í ýü Û ” p o ’ Å ” p o ¢ õo Q   Å ² Æ ³p œ is œ q œ œ Ê Æ p53o ¢ Å É P   ÿ sq p œ ¥£ £ £ £ ¢ q Æ Û ” p o ªÅ ” p o ¢ Ço ’ Å œ œ Ê Q   Æ p53o ¢ Å É P   ÿ sq I í ý (ƒü f p ” ” 5k ” ¢ 1o ’ k Ê U V p o œ   É p53o ¢ Å   p Û ” p oV’ªÅ ” p o ¢ õo Å Æ Ê q ÿ q þ íýƒ1Sƒ €e† œ ü œ œ ß ’ k”¢ k Û œ q Û † ” ” ¥£ £ £ £   ¢ Since the pair of processes is Markov. so that p ’ Å ” ¢ Ço Å É p œ YªÅ ” p o ¢ Å o ’ œ œ f U V p o œ Û is Æ p   53o ¢ Å ß ÿ sq þ í ý üS € ƒ(ƒ1ƒ e† p ¯Û ” o œ Ê œ The value at time Û Èi valid for all I We seek a solution of the form The partial differential equation for is a martingale.

We have (IC) (3) (2) (1) ”p ’ ” o ’ ’ ù ’’ ‘ ¢ q p § ú ’ ”p ” o ’ ’ ’ ’’ ‘ ¢ q p ” o ’ § § ž (h ¢ ¢ ’ o ù o ” b’ p ” o ¢ ƒ’ ‘ ¢ i q p ” o ’ ù ’ ‘ ¢ q p ” ’ ’ 5™ p ” o ¢ ¢ m † ‘ m § § § § § (h † ž ú ú¢ ’ ¢ p o ’ ” o ¢ ’ ’ ‘ ¢ (i q p ” o ’ ’ ’ ‘ ¢ q p ” Y’ ’ ™ p ” o ¢ ’ ¢ † ‘ § § § k § (hl† ž k † ™ —d k p ho §1r— p ”o¡ — p§1o ’ h p ” o ¡’ § p h §1o ¢ § p ” o ¡¢ † § ú ’ ú ú¢ ’ ”p ¢ o ” o ’ ’ ¢ ’’ ‘ ¢ q p ” b’ p ” o ¢ ¢ ’ ‘ ¢ (i q p ” o ’ ¢ ’ ‘ ¢ q p ” o ’ ¢ ’ ‘ § § § § § k ¢ ™ p ” o ¢ ¢ § § h . This implies the initial conditions satisfying (*) ” d We must have † Ê We get three equations: We want to find † ” because 322 corresponds to for .¢ ’ ’ ‘¢ ™ ’ 5k ” ¢ 1o k ¢ ’ ’ ‘¢ ™ ” ¢ Ão ’ k k ™ Ê ¢ ’ ’ ‘¢ ™ ° p q ú k¢ o ú k¢ o ™ ¢ k ¢ k q Ê ™ p ”o ¢ m s ´p ¢ ’ ù ” o ’ ’ ù ’’ ‘ ¢¦™ p ” oV’ p ” o ¢ ’ ‘ ¢ (‰™ p ” o ’ ù ‘ i § § § § p ” Y’ ’ m p o ¢ ¢ m p o ¡ — ° p ” ” o q ” q ” § ú ’ ú § ú ´p ¢ o ” o ’ ’ ’ ’’ ‘ ¢ ™ p ” Y’ p ” o ¢ ’ ’ ‘ ¢ (‰™ p ” o ’ ’ ‘ i § § § § p ” o ’ ’’ p o¢ ’¢ p o ¡ ’ 5p ” ” ’ k ° q ” q ” § ú k § ú ’ § k ú ´p ¢ ” o ’ ’ ¢ ’’ ‘ ¢ ™ p ” oY’ p ” o ¢ ¢ ’ ‘ ¢ (‰™ p ” o ’ ¢ ‘ i § § § § ¢ p ” ’ ¢ ’ o p o ¢ ¢ p o ¡¢ ™ —d q ” q ” § k § k §ú ´p p ’ k ” o ’ ’ ñù ™ 5ƒ’ ™ ¢ ú ú ú ¢ §p p ’ ’ k ” oY’ p ” o ¢ pxù ™ ¤k’ ™ ¢ k ¢ o ’ ‘ ¢ (i‰™ p ” o ’ ñù ™ 5ƒ’ ™ ¢ ‘ § § p ” o ’§ ’ s™ ¤ƒ’ p m ’ k’ § k ¢ p¢ ’¤k ™ p ”o ¡ ¢ k ™ § § ’ k ™ ¤ƒ’ ¢ k ¢ o¡ ™ ¢ k Í¢ o q p ” ‚ — ™ p ” o ¡ ’ k § ” ’ ” k ” 5k ¢ 1o Ê ¢ k † ’ ¢’ ’ ‘ £™ ¢ ’ ¢ ’ ‘ £™ ¢’ o q k ° 5p ” ’ ” k h ” ¤k ¢ 1o Ê † Ê Ê Ê (PDE) becomes ’¤k ” ¢ 1o ¢ ¢ k ‚4 ‚ ’¤k ” ¢ 1o ¢ k ‚ ‚ ’¤k ” ¢ 1o 4 4 k ‚ ‚ ’¤k ” ¢ 1o ¢ k 4 ‚ Ê ’¤k ” ¢ 1o k ‚ Ê ’¤k ” ¢ 1o ˆ k Ê ” ”o ’ p ”o ¢ § § p ” 5k ” ¢ 1o p o ’ ’ k p ”” Ê ” ’ f ” ” p ” ¤k ” ¢ 1o p V’ p o ¢ ë† p ” § k ” ’ ”o ” ” ” p ” 5k Ê ” ¢ 1o § p o ¢ ë† p ” § k ” ’ ” Ê ” ’ ” p ” ¤k ” ¢ Ão p Y’ ë† p ” § k ” ’ ”o ” ” p ” ¤k ” ¢ Ão Ê p o ¢ § q † p ” k ” ’ ” Ê ” ” p ” 5k ” ¢ k1o €p o ¡ — p o ¡ ’ ¤k p o ¡¢ ¢ k § q † p ” ð ’ ” ’ ” Ê ” q ” q ” q § î p †j— ” p h Q¿ § ” ”o h pÃor— §1o ’ h p h p h 1o ¢ f † † § ë† §Û †¢ k k p §h ” ’ k ” 9œ 1o ’ k ” ¢ !ò ” .

which implies that We must choose the parameters so that these three equations are satisfied.CHAPTER 32. 32. and then integrate (3) to obtain the function . A two-factor model (Duffie & Kan) 323 We first solve (1) and (2) simultaneously numerically.3 Calibration Ÿp p oY’ p ªÅ p o ¢ p o ¢ Å o ’ p ” o ’ ” ¢ õo Å P ” oj— P ” p bªÅ p o Ê q P” q P” q § œ § œ š í ýü ¤ƒ¨† œ œ P and the yield is This equation must hold for every value of and . P ú ú ž Y’ ‘ ” i ” ¢ ‘ wù ” ’ ” ¢ b’ m ” ’ ” ¢ ’ ž ž ’ k k ¢ Å Å But we have set up the model so that Thus is the yield at time of a bond maturing at time ” ¥™ œ ” ” h h p o o ” r— P ” p P ” ’ f † † § p ªÅ o ’ p o¢ Å œ œ p j— o o ’ ™ p P ” Y’ p o ªÅ ™ p P ” o ¢ p o f ­ P” § œ § œ p o ’ œ œ P ” p j— o o ’ p ™ p P ” Y’ p o ªÅ ™ p P ” o ¢ p o ¢ Å f ­ P” § œ § œ © d † P † ” o ” Å ” p V’ Å p o ¢ õo Ê œ œ ¢ ž ¢ ” ’ ¢ ” ¢ p P”o ¢ © d m P ” k p ªÅ o ’ † œ § k   xo P ” q Þ d h Q¿ P Let be given. . The value at time of a bond maturing at time ” ¥™ œ œ p ” j— o ” is .

324 .

which we call a “stock” even though in applications it will usually be an interest rate dependent claim. is a Brownian motion relative to this filtration. Ÿ – ëi Û œ i  ƒp o h ž œ ¥ š – Û ”p o œ œ p o ‘ C ” Æ p o p o u™ p o ‘ Æ Q R  œ Æ “ p¤3o É P    ÿ œ œ I ) C p o í ýü (ƒë† œ œ “ p o  œ p o p o  † œ Ê ² Æ Ï…Î ú œ “ Æ p o œ  C (0. of return for the stock is We define the accumulation factor The zero-coupon bond prices are given by 325 U V p o œ ¥£ £ £ £ ¢   Æ p   53o ” ´p o  f p o ß œ ɜ ÿ sq C ¥£ £ £ £ Æ ú p¯õo Û ú ° p o  e† € œ þ ƒ1ƒ e† í ý üS € p o p o ‘ p œo œ œ “ ú † p ¯Û ” o ³ p o p œo œ œ ú Ð Ñ “ ÑÉÉ Ï Ð so that the discounted stock price is a martingale. We shall work only under the risk-neutral measure. This is not a geometric Brownian motion model.Chapter 33 Change of num´ raire e Consider a Brownian motion driven market model with time horizon . For now.1) Ÿ – ëi Û œ i h ƒp o ž œ ¥ š . which is reflected by the fact that the mean rate . but may be larger than the filtration generated by . Indeed. given by a term structure model we have not yet specified. we will have one asset. We are particularly interested in the case that the interest rate is stochastic. The price of the stock is modeled by where the interest rate process and the volatility process are adapted to some filtration .

the e e stock is worth units of money market and the -maturity bond is worth units of money market. so Definition 33. œ œ W a q ´p o “ ¥£ œ œ £ £ £ p Û ¯õo † ú d Û ° p — ¯¡o †  e† € p ¯Û ” o ð ñH p ¯Û ” o € œ ð H p o p ¯Û œ ” o œ 㠀 W W  œ ú Ê î ã í Ü ë ï–ìeê  ú Û Û . We then denominate all other assets in units of this num´ raire.e. the price process for some asset whose price is always e strictly positive. i. At time . ã é ¥Ü Example 33. ´p o œ œ ¥£ £ £ £ p tÛ ” o ¯Û ” o p p o Ê W ‰q ú œ œ p o œ ú † ptÛ ” o p¯Û ” o p o “ Ê W ‰q p œ o ú œ œ ££ Û ú œ † p Û ¯õo ú ´ £ p¯õo œ ú ° ¯Û ” o ° “ p o p £ p o p W‰q W ¯ÛÇo ú € ¥ ££ œ œ † ptõo € œ Û œ ´ £ ú ¤° p o £RtÛ ” o pp ptõo o Û h p“ o  e† € œ is also a martingale (tower property).1 (Num´ raire) Any asset in the model whose price is always strictly positive can be e taken as the num´ raire. e divided by . is a martingale under . The value of the forward contract at time is zero. At time e e is worth units of -maturity bond and the -maturity bond is worth 1 unit.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 .1).2 (Bond as num´ raire) The -maturity bond could be the num´ raire. the stock . e e Example 33. Then defined by  H  is risk-neutral for . The original probability measure is risk-neutral for the num´ raire (Example 33. ã ð ñH 㠀 ” p – õo Û ¥ ¸ — ½™ò ” H € ú p – õo Û Ûo Æ p – õu ˜ ÿ p ho §1u f € p¯Û ” o Ê p oœ ð H d € œ “ Therefore.71 Let be a num´ raire.1 (Money market as num´ raire) The money market could be the num´ raire.. Ü á ß gàÞ á æ ß Þâ èç 3åä ã á ß gàÞ á ß ŽàÞ âÝ We will say that a probability measure is risk-neutral for the num´ raire if every asset price. e Theorem 0.

have the same probability zero sets.54). and is the price process for some asset. is a martingale under . is a probability measure. Under . The risk-neutral measure for this num´ raire is e e  Ô under . be the num´ raire. is a martingale. then  # Ô H € ” d ú p h §1o Ão  † p h p ho 1u d † – Û ú p h Ão  ´ p õo ° p – õu Ûo f d ú E € † p – õo Û p h Ão  p E  3o Ûo Æ p – õu ÿ d ú † #  ð H € ð H € p — t3o † H H € ¥ f p – õo Û ¸ — ½™ò ” ð H € p – õu Ûo ú Û Æ p – õo ˜ ‹ ÿ p ho §1u € ð ñH Proof: Because  and we see that € H € Note: CHAPTER 33. ð H € 33. Therefore. is a martingale. We must show that under . 327 ¥£ £ p¯Ço  Û ° p Ûo tõsÔ € Ô .. For this. ý ôö  ¶’õ  ÿþ ¢   ÿþ ý ô ¶Žö ý ô ¶Žö ý ô ¶Gø ¥ ô Žö © © û ý ô ø ÷ ö G3©û §¦ ý ô ø ÷ ö G3©û ¥ ¥ £ £ ¤ý   ¢ ¡ÿ þ Fix and let ð ó ý ô ø ü ö ¶h–Ž©û ú ù ô ø ÷ ö õ v¶h3Xrô ö § ¨¦ ð H ” € ¥£ £ £ £ ´p o œ ú p o p œ ‡Ô o ú ptõu ¯Ço œ Ûo p Û ° p Ûo p Ûo ¯õsÔ tõu € p o  p œ sÔ o p œo  ú pœ o p œo  ú œp o œ † † † ¥£ £ £ £ ´p o œ ¥£ ú £ Û p o  £ ´ p o £qÅ p¯Ço ° ú p Û tõou pœ o f ¥£ œ ú € † £ p¯Ço 1ou Û p h p uœ  o ´ p o ££ Å ° ° Å ú p¯Ûõu o p h ¥ p o œ Ão  € † œ ð ò € œ Å – ÈÀui jh Û i Û i p Û ¯Ço Let be an asset price. i. we need to recall how to combine conditional expectations with change of measure (Lemma 1. Change of num´ raire e and are equivalent.e.1 Bond price as num´ raire e # ´p o £ £ œ ú # Ô which is the martingale property for œ H € Therefore. If and is -measurable.

in (2. We may assume without loss of generality that ü 1 ü 1 35A÷  ô 1 ü 1 45352÷ ø ý ü ‹iŽö   C¤  @ ' ø ý ô ø ü ¶h–Žö ý ý ü e’ö ô ø ü G–’ö 0 ) û ý ô ø ü ö G–’¶ £ ý ôøü ¤¶h–Žö 8 6 £ ý ôøü 97¤¶G$Žö )   5ÿ þ  ) )  ô is called the stock is -forward measure. In terms of this num´ raire.1) can be rewritten as ¥ ) Q ` T ÷ G ý ô ø ü ö s¤¶G$ŽïT Eù 1 ü 1 X¶ô 2ÔW÷ where generality that ý ü e’ö is a Brownian motion under . i. the value of the -maturity bond is This is a martingale under . ô ”. a differential without a term..1) in (1.1’) . we change the measure only “up to time using and only for . the value of the (1.. We may assume without loss of . (2. the stock price is identically 1. In other (2.1) is a Brownian motion under . (2.72 The volatility words. This is a martingale under . and so has a differential of the form ø ô ¶5 33.e.1). and so has a differential of the form ý ô ø ü ö ¶G$Ž¶ 8 Y6 ø ý ü ‹iŽö H @ ' S ý ü iŽö ¥ ) Q ý ô ø ü ö ¶G$Žó 8 6 $a£ S ý ý ü iŽö ô ø ü G–’ö Theorem 2. The process .328 ô Because this bond is not defined after time . The riske e neutral measure under this num´ raire is e ý ü e’ö Denominated in shares of stock.e.1) is equal to the volatility  ý ù ô ö Ž£ ø ý ü ‹eŽö  (õ H HIÿ þ @ S ýü V ¨eŽö ¢   ý ü e’ö ø ¥ ) ÿþ ¥ )  Q ý ù ý ù £ ý ô ø ü ö ¶h–Žï ô ô Žö ô Žö ý ô ø ü ¶G$Žö ý ü iŽö ©0 T U£ §¦ 0 û ý ÷ x3ö ¥ S ýü ¨e’ö 0 ¥ ) £ ¤ý ý Q R ô ø ü G–’ö H ¢ Iÿ þ ö H Pÿ þ C ý ü e’ö We write rather than from now on. ÷ G ý ô ø ü ö 8 ©3¶h–Žó$6 @ E ô 1 ü 1 FÑDÑ(÷ B i. Denominated in units of ô ý ô ö ¶Ž¥  (õ ¢ #  # "  %  '& $   ! #  %  "    H ) @ B 0   ¡ÿ þ   ¡ÿ þ -maturity bond.2 Stock price as num´ raire e Let be the num´ raire.

e.3 Merton option pricing formula v ýü 5eŽö   @ i £ ýü ‰ˆe’ö H H @ @ ý ô ø ü ö ¶G$ŽÈ e X¥ Proof: Let CHAPTER 33. in (2. Then 329 .ý ÷ x¡ö ý ÷ x¡ö ý ÷ x¡ö ý ÷ x¡ö 0 0 0 £ 0 £ £ £ £  ý ÿþ ô Žö © ý ô ø ÷ ö ¶h3©û ¥ p q• A¶Žö o ý ô VElyk3¶’ö B ý¶Gø¡ö ûŸ•ri • ý ô ô ÷   5ÿ þ n H FIÿ þ ¥ ¥ ) E • ý lyk3¶’)ô ö B ¶ôG¡ö Ÿ•WmEly3¶Žö B ý ø ÷ û i • k ý ô   H 0 5ÿ þ 0 Iÿ þ Elyý¶Žö • k ô ý ô ÷ • k ý ô ) B 5ÿ þ ¶Gø¡ö ûŸWimEl•y¶Žö ) B Iÿ þ   Hd 0 d ô ý  ý Žö x÷¡ö h f e# xgh%   H ¦ h f e# xgd$   H ˜ ¦ ý ô ø ÷ ¶Gè3ö û Ÿ • Wi ô Žö ÿþ ý ô ¶Žö © 0 d ™ giý h f e# jgh%   H ¥ © “ ’þ • Wi ˜ d ™ 9'ý 0 ö ô Žö h f e# xgh%   H ý ô ¶Žö ý ô ¶Žö 0© © “ ’þ “ ”’ þ ˜ –ý • i ý ô $—ŸWI¶Žö ¥ £ ý ÷ ¤x¡ö ‘  5  ý ô ø ¶Gñ ö  d 8 6 … ¦ Under is a Brownian motion. but the change of measure does not affect the volatility. has volatility and mean rate of return . The change of measure from to makes a martingale. Change of num´ raire e The price at time zero of a European call is 8 ‡6 ý ô ø ü ö ¶h–Žï T ý ô ø ü G–’ö r8 %6   @ i Vø   5ÿ þ ý ü eŽö Q  ¥ ) c £ ý cö ¤d1b . so . Therefore. it changes the mean return to zero. i..1) must be and must be ý ô ø ü ö ¶G$Ž¶ 8 Y6 … # † 8 H Pÿ þ 8 … # †   5ÿ þ ü  ý ô ø ü ¶G$Žö r ) ý ô ø ü ¶G$Žö  @' ‚ƒS iŽö Q ý ü ú „¡ü ý¶hø–Ž¶Y6y¡eŽö ô ü ö 8 v ýü i ý ô ø ü ö 8 6 ¶h–Žó$a£   ¥ )  @ V w ýü 2iŽö € ý øü ü ¶ôG$Žö r8 6y5ýiŽö v ü ý ô ø ö 8 ¶h–üŽ¶Y6xi £   ¥ ) ý ü eŽö t ýe’ö r ü @ V v5e’ö ý ü ý¶G$üŽö ô ø ý¶h–Žöó$6 ô ø ü 8 i p£ r8 6   ¥ ) r   ¥ ) ýeŽö ü ý ü iŽö ý ý heüŽö ö ff b  ) 5ýeŽö v ü ý ý GiüŽö ö f u£ b ) ) ) ) ) b £ S ý ý ü Ge’ö ö ñ ) t s £ ý cö f b ø r gc a¤h¡f g3¤9c p¤h’gb i £ ý cöf e e q¥ 33. Under this measure.

then ý ô ¶Žö û ý 3ö ÷ ‰ ˆ‡ £ ý ôø ÷ g¨7¤¶G¡ö 8 Y6 £ r ‚ † 0 û Ÿ ~ } • „!{ “ ô   8 Y6 ý ÷ x3ö )  ¥ 8 6 £  ‚ where Similarly. When formula • ý 3ö ÷ 0 • 0 ~ } i†{ “ ~ } i†{ “ ô  ô 8 $6  ¥ $6 8 ¥    ¥ £ r ‚ £  ‚ If is constant. 0 û Ÿ • „!{ ~ } “ ô Q @ H 8 Y6 ý ô ø ÷ ¶Gè3ö û £ ¥ ) H Iÿ þ ¥ ) ý ô ø ü G–’ö 8 6 8 6 where 330 . we still have the explicit 0 £ ý ÷ Px¡ö ‘ ø ó r ˜ ô 3iý r8 6 pxhrv  i †ö r ˜ ô 3iý r8 6 ”ˆhrv  v †ö ø ó r  ˜ x3ô r8 6 …i  ý ô ø ÷ ¶Gè3ö ý ÷ x3ö p ˜ ô r r8 6  i r ˜ 3ô r8 6  p p ý r ‚ö  ø ‹ý ƒ5a£ 8 ô ý¶G¡ö Ÿ• i†{ ô ø ÷ û ~ } ô $6 o 0  @ i ¡ÿ þ £  ý ÷ x3ö ý¶’ö ô “ n   ¥  ô 0 ~ } i†{ ô 8Y6 ý ÷ x3ö v k  @ ¡ÿ þ £  ý ô ¶’ö ý¶G¡ö Ÿ• ô ø ÷ û “ n   ¥ 0   ~ i}†{ r ý x÷3ö @ k…ô r8 6  iPý Žö ô 8 $6 £ E • k ý ô Fly3¶Žö B ô ø ÷ G3ö©Ÿ• û       5ÿ þ n ¡ÿ þ ) v us ý ôø ÷ @ w g¡¶Gè3ö û 0 ¶Žö £ ý ô ø y ô r i ý ô r8 6  I¶Žö  ˜ ô r r8 6  v ý ô ø ÷ ¶Gè3ö ý ÷ x3ö p ô  ‚ö  ø ‹ý ƒ5a£ ô r ý¶h3ö ûŸ• ~i}†{ ô 8$6 o ô ø ÷ £ 0 8 6 €v   @ ý ÷ x3ö ý ô ¶Žö n zH ÿ þ  ¥ H ~ } S • o p ¶hø3ö ûŸ• i†{|o 8 r ý ô ÷ @ 0 ô r 6  iIý¶Žö ô 8 $6 £ ý 3ö ÷ H H n Iÿ þ ¥ 0 w v us ý ÷ xgtx3ö ý ô ¶Žö ø e r i ý ô y zô r8 6  I¶Žö This is a completely general formula which permits computation as soon as we specify we assume that is a constant . If . ‚ö ý ôø ÷ö û • i ‚ ý ÷ ý r ƒ5—¶h3©¥Œ¤ý  ‹ö Šx3ö † is not constant. . we have the following: ý ô ø ü ö ¶G$Ž¶ 8 Y6 and we have the usual Black-Scholes formula.

hold shares of asset . hold shares of stock. Change of num´ raire e ô ” 331 . Let At time . At each time .73 Changes of num´ raire affect portfolio values in the way you would expect. ý ý ü Ge’ö We want to verify that this hedge is self-financing. this question takes the following form. will for ? Formulas (3. so we simplify them by a change of num´ raire.CHAPTER 33. then for r ƒ5xi ‚ö  • ‘  ‚ r ‚ £ I™ ™  ý ü e’ö – ‘  ‹ö  ‚ ý ý ü GeŽö ý ü e’ö ü ô shares of stock and short r ƒ5• ‚ö  0 ý ü eŽö ý ü iŽö ˜ A‘ ‘ ü (3. and let be the value of the portfolio at time . . then will be invested in the bond. The number of shares of asset held at time is ü i ýüö 3e’”’   ýýü ‚ö GeŽö r ƒ5 ý ü iŽö 0 – # †ü…  ü #%•ˆ!  “  ” "h… † #… ý ô ø ü ¶G$Žö ‘ — 50 ý ôøü ¶G$Žö ø ¼  ‚ö ý ý ü GeŽö ƒ5 ü 8 ~ } ü i r • i†{ Xuô ‡6  ýü i ô m˜„eŸyŽö r8 6  i £ ýü eŽö  ý ü eŽö “ ¥ 8 ~ } ü i r • ) i†{ Xuô ‡6 £ ýü eŽö ø¶„eüŸiyŽö r8 6 ”v ˜ý ô  ý eüŽö “ ¥ ) 0 • i ýýü ‚ ýü W¤GiŽö  ‹ö ŠeŽö ˆiŽö £ ýü r ‚ö  ûŸ•Wi¤ýheüŽö ƒ5 ý  ‹ö   ‚ 0 0  qXqø r ø  ø û r ‹ö Š¶G$Žö ‚ ý ôøü ‘ ¡ýü ¢iŽö  ý ý ü Ge’ö ý ü eŽö 0 £ ýýü ‚ö ý ôø üö û • i ýýü ‚ö ýü ¤heŽö r ƒ5—¶h–Ž©ŸW¤Ge’ö  ƒ5—iŽö ˜ 0 0 ý ü e’ö ý ü eŽö  ýýü ‚ö  £ ýü GiŽö  ƒ5a¤e’ö ˜ A‘ 1 ü 1 5Ž÷ ‘ û Ÿ ý ô ø ü ö ¶h–Ž©û — 0 ýiŽö ü — œ £ ýü ˆiŽö  ‘  ˆ  —˜ WPiŽ©’ Ÿ ž i ýüö i ýüö IeŽ©’ ý ü ö e’” ý ü ö eŽ© v ýü $eŽö ¥– ’ ’ ‘ 0 V i –ø ø s A›Xqqèšø ý ü eŽö ‘ ý ý ü GiŽö ý ÷ ö x3© ü £ ýüö PiŽ©’ 8 $6 ¥ ’ As this formula suggests. the value of the portfolio will always be the bond always be ý 3ö ÷ ø ý ü ‹eŽö and we will have a hedge. the value of a European call expiring ø ý ý ü ‹Ge’ö where This formula also suggests a hedge: at each time . so the number of bonds owned is and the portfolio value evolves according to The value of the option evolves according to ý ü e’ö ô 1 ü 1 ”5A÷ ý ü e’ö £ ýüö ¤iŽ©’ ý 3ö ÷ £ ý ÷ö Ix3©’ ‘  ‘ ý¶ôG$Žö û ø ü • i ý ôøü ŒI¶G$Žö û Ge’ö r ‹ö Œi ý ý ü ‚  • 0 '  0 0 V  ‹ö 7¤eŽö  v ýýü ‚ 5Ge’ö  ‹ö  e’ö $eŽö ý ü v ýü ý ý ü heŽö ‚  £ ýü ‘ If .1) (3. Suppose we begin with $ and at each time hold shares of stock. Mathematically.2) . hold bonds. If is the value of the portfolio at time .1) and (3. and invest the remaining wealth in asset . Begin with a nonrandom initial wealth . e Proof: Suppose we have a model with assets with prices . Theorem 3. Will the position in ? If so. e This change is justified by the following theorem.2) are difficult to compare. if at time is is constant. We short bonds as necessary to finance this.

 0 ¬ ˜ ˜ ‘—   e 0 ¡³  Ÿ— e ˜ ˜¥‘ ˆ  ˜  0  ˜ A‘ ˜ ¡ S ˆS  Q  ' S  Q  0 0 v v ˜ ˜ ¥  ¥  S  Q` V 0 Ÿ ˜ v S  Q` 0 Ÿ  ˜ ¦ ˜ ˜A‘ — ¤ © ˜ ˜¥‘ — ¤ S   ¥ ¥ S  Q  ¥  –ø ø s $šXq›ø ¥ £ 5™ ø ’ v v ˜ Q   0  Ÿ  ˜ £  ¬ ˜ ¥‘ — ¤ ˜  0'  Q Ÿ  ˜ £ ˜ ˜ ¥‘ — ¤ ¥  0 ' Ÿ °˜  £ ˜A‘ — ¤ ¥  ’ v Œ¯’  ¥ ¥ £ ®’ «  Then ý ü ö iŽ5 ý ü ö eŽ©  ý ü eŽö ýeŽö5 ü 0 0 £ ýü ¬ §eŽö ­ø ˜ ˜ ’ £ ýüö ¤eŽ5’ « Let and 332  ø ý ü ‹iŽö ˜ 0 ý ü e’ö —  ˆ — qXqø  ø    ‘ ‘  ‘ Ÿ £ ýü  ˜ eŽö — ’ ˜ ‘ —¤ . ’ .— 0   ¬ 0 Ÿ i  ˜ ¨’ ¦ v 0¬  Ÿ £  ˜ ®’  — 0  ¬ ¬ ƒ˜ ˜ ‘  ˆ ¤ — © ˜ ˜ ‘ —¤ « « — 0 ¬  £  ˆ °—˜ žWi¨’ 0 Ÿ « œ ±£ ž i W² ’ e — 50 ˆ °—˜ W£’ Ÿ ž i œ ±£ — ‘ Now Therefore. not — 0 — ' 0 0 V  ˜ A‘ ˜  0 0 Ÿ i ’ ¦  ˜ ¨'§v ' © ˜ ª˜ A‘  ˆ ¤ — ˜ ¥‘ ˜  Ÿ  ˜ —¤ £  Ÿ £  ˜ £’   ˆ ¤— and we only get to specify Note that be a num´ raire. and define e evolves according to the equation . in advance.

the option value formula e ý ý ü GiŽö becomes This is a homework problem.1) and (3. the asset values are e ø ý ü ‹iŽö The portfolio value evolves according to In the new num´ raire.CHAPTER 33. Change of num´ raire e 333 .1’) (3. and all assets and the portfolio are denominated in units of shares of asset . We return to the European call hedging problem (comparison of (3. but we now use the zero-coupon bond as num´ raire. we must show that   ýýü ‚ö GiŽö r ƒ5 W¤iŽö • i ýü )  ý ý ü GeŽö  ƒ5  $GiŽö  ƒ5  ‚ö v ýýü ‚ö  ý ý ü Ge’ö  ‹ö   5GiŽö  ƒ5  ‚ v ýýü ‚ö ý ü eŽö ) v ýü 5eŽö ) and ý ý ü GiŽö ýü eŽö ) ø ý ý ü gGeŽö  ý ü e’ö r ƒ5r´eŽö ‚ö  • i ýü ‘ r ƒ5—¶G$Žö ‚ö ý ôø ü £ ý ¥ ¥ iGe’ö  ýýü ö  û Ÿ £ Bond: ¥ £ ) ý ý ü heŽö ‘ • i ýü ŒPiŽö ý ô ø ü ö ¶h–Ž©û ý ô ø ü ö ¶h–Ž©û ý ô ø ü ö ¶h–Ž©û i ýü ö v ýü ¤e’ö ’ l5iŽö ý ô ø ü ö ¶h–Ž©û  ‚ ƒö5a£ £ ýü ¤e’ö ý ü e’ö «‘ ‘ 0  ƒ5a¤iŽö ‚ö  £ ýü ‘ 0 Stock: (3. In terms of the new num´ raire.2)).   ÷ £ ýýü ‚ö gl¤GiŽö r ƒ5 r´eŽö • i ýü ) To show that the hedge works. We still hold e shares of stock at each time .2’) ü i –ø ø s P›qXšø £ 5™ ™ ý ý ü GiŽö )  ƒ5piŽö ‚ö  £ ýü  This is the formula for the evolution of a portfolio which holds . ¥ ˜ A‘ ý ü eŽö ‘ « ý ý ü GeŽö )  ý ü e’ö ‘ £ ýü ¤e’ö ’ «  ý ü eŽö  ƒ5a£ ‚ö  )  «‘  ¥ .

334 .

The bond prices ô ø ü G–’ö ý ü ø ü eG$Žö û ¸ ø ý ü ‹e’ö @ '   ý¼šø$Žö6 … P¶h–Žö ù 6 ü £ ý ôøü   ¦ ý ô ø ü ö G–’ï 6 v |5ü ü  ý ô ø ü ¶G$Žö ù 34. This is not possible because it leads to volatility of 335  › ¸ ø  …     ý ø ü 6 £ ý ôøü ¦ ¶ôG$Žö r aI¶G$Žö ¸ ¸ ¦ … aI¶G$Žö ù 6 ý›–üŽö ø 6 £ ý ôøü ¸ ¦   ÷ © k W6 .1 Review of HJM under risk-neutral Brace-Gatarek-Musiela model Chapter 34 Forward rate at time for borrowing at time  zô 6ý ôø üö Š¶G$Žï6 £¤¶h–Žö ý ô ø ü ¸ £ ý ôøü ¤¶h–Žö ¸  ·¶µ The interest rate is £ ýüö ˆiŽ† where A simple choice we would like to use is To implement HJM. you specify a function where is the constant “volatility of the forward rate”.ø ý ü ‹iŽö @ ' ý ô ø ü ¶h–Žö ¸ 6 v y$ü  ©   ý ø ü  –Žö ý ô ø ü G–’ö ¸ 6 £ a¤ý ô ø ü G–’ö 6  ô 1 ü 1 45352÷ @ ' -maturity bond. ø ý ô ø ü ö ‹¶h–Žï 6 ýü eŽö ý ô ø ü ¶G$Žö û ý ô ø ü ¶G$Žö ù à  ›Á   6 À i Pü  ý ô ø ü ¶G$Žö û ý ü ö iŽ † £ ý ôø üö a¤¶h–Ž©û  satisfy º v u ¿gs  » ý øü P ›–Žö … i £ ¸ ¦   ¼¼ º v u zgs  ½ ü ¼ ¾ýe’ö ¼³»P ýhö|† … i  £ ¤ý ¹ ²’ þ   ¦ .

2) ô Å ýšø$Žxip¤|5G–’ö û Å üö † £ ý Å v üø ü £ ý Åøü Æ Iš$Žö Æ Æ v u gÆ s Ç p   ý›–üŽ† Ç ø ö £ È i n ¦ p  v u gs    ý¼y%hø–Žö È i  v ü ü £ 3zÉ 2 Ê ý £ C ¸ ¦ n v u gs p  É ý›–Žö ÈY– … i Éøü £ ¸ n … ¦ ý Å v ü ø ü |5G–’ö lIš$Žö û £ ý Åøü ü X i É £  ¿a2hö . To see what problem this causes. Jarrow and Morton show that solutions to this equation explode before .ý Åøü š$Žö ø ý ü ö ‹iŽ † £ ýüøü a¤eh–Žö ¸ £ ý ÷ ø üö Ix–$Ž† Ç ô Å ýÅ|v$G$Žö ü ø ü £ ý Å ø üö Æ P›–Ž† Æ ¸ Æ Æ Å v üøü y$G$Žö P›–Ž† £ ý Å ø üö ¸ ø ý ‹ e q¥ ø ¥ Ä £ ˆÄ i ü Iq£ e q¥ i ü I$Ä ü q ý ÷ x3ö ü$ÄWi ¥ ˆiŽö  £ ýü Ä ¸ ý ü iŽö i ü ¸ Iq£ ¸ i £ Ž ¥ ¥ … ¦ ø £ £ £ ý ÷ x3ö ¥ ý ü eŽö ¥ ø ý ü iŽö ¸ v ¸ i ¥  ¥ ü ¸  i f ¸ ¸ ¸ where ÷ s ý ü iŽö ¥r ¥ ý ü ö e’’ ø ý ü geŽö r ¸ £ ýü ¤iŽö f ¸ ü The problem with the above equation is that the term grows like the square of the forward rate.3) (2.1) (2. consider the similar deterministic ordinary differential equation  and Heath. Current time ü Time to maturity ü i ô £ šŸy‘2Å Bond prices: Forward rates: (2.4) (2.2 Brace-Gatarek-Musiela model 336 . ô This solution explodes at Ä £ ˆü k Ä £ ý ÷ Fa¤x3ö New variables: 34. We have .

7) (2.6) We have . (2.8) differential applies only to first argument year). .(2. In this notation. the HJM model is (2.5) and (2.3 LIBOR (2.(2. analogous to (2. $ invested at time in a -maturity bond grows to $ 1 at is defined to be the corresponding rate of simple interest: ý ¤ Ì v ü y%’ö ü ý ø ü ¤ $Žö Ç Ì š  Í zÌ £ Î |5ü Ì v ÷ q2Ì k .6) (2.6).10) (2.Ì   Ð v u £ ÷øüö gw gs ¤ý ––’¶Î i y  ¼š–’|† ý  ø ü ö ¥ Ï º v u ¿gs ¤š$Žö  ý Ì ø ü ø©»I ý¼šø–’|† ü ö £ £ ý ÷øü ÎÌ Px$–Žö ¨yv ¥ ϳ¦ ¥ Ç £ ýý ÷ ø üö Î Ì Phx–$Ž¶¨|v 8¤š$Žö ö ý Ì ø ü ¥ ø ¥ ý ÷ ø ü ö x–$Ž¶ ýü eŽö  ý ø ü  –Žö @ V @ ' ü ý ø ü  $Žö ý  Å š ý ø ü  $Žö ù Å š 6 i rPü Å ›  ý ø ü  –’ö Å š ú ý ø ü ö V $Ž ý  Å š † i ý ÷ ø üö r¡x–$Ž† ‚ £ £ ü Å › ý ø ü ö  $Ž Å š † i ýü W¤e’ö Ç  ý  Å v üø üö u%h–Ž©û ô Æ Æ v ý  à u%h–Ž©û  À  Šv üšøÁ ü ö ü  ý ø ü ö  $Ž ü  ý  Å v üøü u9G–’ö ¸ ô Æ Æ v ý ø ü  $Žö Ç (2.9) (2.8) ø ñ  ý Å ø üö 6 £ ý Å ø ü ›–Ža¤›–Žö ù 6 ý ø ü ö ¼ $Ž    š 6 È ¦ ø ý ü ‹iŽö We will now write @V   ý¶ôG$Žö ûhšø–’ö ù 6WiPü ø ü ý Å ü ý¶Gø$üŽö heŽa¤¶h–Ž©û ô û ý ü ö † £ ý ôø üö @ V   ø ü gýeŽö ý Å ü ö šø$Žï6|5ü v ýÅ›–Žö ù —ý›ø–Žöa¤¶h–Žö ø ü 6 Å ü 6 £ ý ôøü ¸ i ôø üö 6 £ ý Å ø üö Ëh–Ž°aP›–Ž°6 rather than ý ô ø ü ö ¶h–Žï 6 ý eü We now derive the differentials of where CHAPTER 34.2) Å š r w ŒÅ @ '  € r G›–Žö ù dö  %›–Ž† Æ £ eüŽö ý ý Å ø ü 6 š–’ö u%ü v ý ý Å ø ü 6 v ý Å ø üö Æ Å @ V  † Æ ¡eŽö v ýü ý Å ø ü ö 6 v š$Žï|$ü ý Å ø ü ›–Žö ù —š$Žï6 6ý Å ø üö £ £ ý Å ø üö Iš$Ž†  Å Æ Æ Also.5).5) 337 .4) Å v üøü |$Ç G–’ö û ý ø ü G –Žö ù 6 i WPü  Å v üø üö û ýüö Ç |5h–Ž©eŽ† 34. (2. Brace-Gatarek-Musiela model and Å š differential applies only to first argument Æ ý  à |%šÁ –Žö  À Å v  hø ü ü ¸ † ý ø ü ö  $Ž Å š £ ý Åøü ¤›–Žö ù 6 (2.1) Ç £ ý Åøü Iš$Žö Ç  Fix time (say.

is the simple rate over a period of duration .ý ü eŽö @ V ý ø ü ö  $Žï  –Žö ý ø ü Å š T Å 5 › ý ø ü ö  –’¶ Å š Î .5 The dynamics of Ì ý ø ü ö  –Ž Ý Ü Û ÚÙ 3'V›WØ We want to choose   Îyv$ü Xq¢ö ¤š–’¶Î ý  £ ý Åøüö G Å ø ÷ G ü ø ý Å ø üö (¤sR‹›–Žï6 ÷ © . At time . ý ø ü ö  $Ž Å š † ø ý Å ø ü ö † ‹›–Ža£ Ò ¼¼ È Dgs º v u  Ÿ ³»I ý¼šø–’ö|† ¼¼ ü ý Ì ¤um›–Ža£ v Å ø üö † Ò – $qÈ ¦ Ò ¼¼ È zgs Ì º v u  Ÿ Ô»P ý¼šø–’ö|† ¼¼ ü Æ Ò – $qÈ ¦ Æ Connection with forward rates: Ì  È Ð w gs Iš–’¶Î v u £ ý Åøüö  i y  ¼š–’|† $qÈ ý  ø ü ö Ò – ¥  ø »  ý  ø ü ö ¼š–’|† Ò – $qÈ  Ð ý  ø ü ö y  ›–Ž† $qÈ Ò –  Ð E ý  ø üö ‡ ›–Ž† È È º gs v u £ ¦ v u w gs i ý ¤Ìy4š$Žö v Åøü vgs £ u £ ý Åøü ÎÌ P›–Žö Óuv ¥ iVB ý ø ü Å›–Žö Ç ý ø ü Åš–’ö ø £IýGš$Žö¶Î¨Ì| ý Å ø ü Ç v ¥ö ¥ ý¤Ìuvmś–Žö ø ü Ç Å › ÇÎ ý ø ü  –Žö  ÷ £ ýÌ v Åøü gÒ¤D|F›–Žö ü v ý Åøü $š$Žö Å v y5ü ý ø ü Åš$Žö i ýDÌ|vFś–Žö ø ü Ç Ì v ÅÇ v uF|5ü ü ü # Ç Ç È Ñ Ò – d… $qȆ Ñ " È # Ñ d† "… Ò # – d… $q# Ȇ Ñ " "… d† 34. Ì ý ø ü ö  –’¶ Å š Î 34. 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. appearing in (2.4 Forward LIBOR 338 is still fixed. Can do this at time by shorting bonds maturing at time . we can have a log-normal model for .2) (4. agree to invest $ at time Å v yPü Ìyv¿Åy5ü v k `Ì ÷ r so The forward LIBOR is defined to be the simple (forward) interest rate for this investment: .1) Å › † . with payback of $1 at time and going long fixed  ÷ © k ¿Ì ø ¥ i y 25  Ì  š ¥ Ì È Ð Fgs Iš$Ž¶Î w v u £ ý Å ø üö  i2y5 ý¼šø$Žö† $qÈ ü Ò –  ¾!ÒÕ×Ö { † $qÈ Ò – È Ð Fgs w v u ý ø ü ö ¼ $Ž £ ý Å ø üö † £ ý Å v üø ü ¤›–Ža¤u%G$Žö ¸ is the continuously compounded rate.5) so that (4.1. For fixed positive .

for some corresponding to particular choices of .9): ý ø ü ö  $Ž Å š 6 . and is a subclass of HJM models.1) ý Å ø ü  £ ›–Žö Î Þ # Í  339 (5. CHAPTER 34. This is the BGM model. (5.1)  .ü  ü  ú r ý ý ø ü h –’ö ù Å š 6 hö r ©  ‚ p @ '  e’ö ý ü ú ý Å ø ü eš–’ö ù Œ¤¤|m›–Žö ù 6 6 i ýÌ v Åøü v £ ‚ ú ý Å ø ü V›–Žö ù rI¤y4š$Žö ù 6 ¤|Fš$Žö ù |v 6 i ýÌ v Åøü ý Ì v Å ø ü 6 ‚ ‚ Ì  ü ú ý Ì ø ü ö † i ý Ì v Å ø ü ö V¤š$ŽW¤¤|Fš–’|† ú ý Å ø ü ö Î Ì eš–’¶¨|v £ ¥ n ¥ p  ‚ r ü ý Å ø ü 6 i ýÌ v Åøü r úVš$Žö ù rI¤y4š–’ö ù 6  v ‚ @ ' ý ü e’ö ú ý Å ø Vš$üŽö ù 6W¤¤Ì|Fš$Žö ù 6 i ý v Åøü v r r ‚ …i r G¤Ìy4Ś–’ö ù 6hö €v$D›ø–Ž†ŒP¤um›–Ž† ý ý v øü  ý Ì üö i ý Ì v Å ø üö ä n ‚ Ì ä ý Å ø üö Î Ì $úVš$Ž¶¨uv £ ¥ È  È ºzgu4„s ¥ v s Ì   ý ø ü ¼š–’ö|† » ý ø ö v ¦ P ¼š–ü’|† Ò$qÈ – Ò$qÈ – ¦ ¦ ¥ È  È zg4Ì º v us    ý›–Žö† øü » ý øü P ›–Žö† £ Ò – %qÈ Ò$qÈ – ¦ ¦ ¥ ß l È Ð Fgs à w v u âã ¥ i y ᡐ  Ì ý ø ü ö ¼ $Ž  š † $qÈ Ò – ý ü eŽö ‚ @' ý eü’ö ú Å ø ü Vý›–Žö ù 6Wi¤¤|vFš–’ö ù 6 ý Ì Åøü v r r  w ü ý ý Å ø ü 6 ý Ì Åø 6 v ý Åø ü † ý v Åøüö € r Gš$Žö ù dö  i r Gý¤yv4š$üŽö ù dö  $š$ŽöWi¤¤Ìy4š–’|† £ È È r @'    w   ý¼šø$Žï6 $–qÈ  üö Ò v%ü  € r G›–üŽö ù dö ”$¼šø–’|† Æ $qÈ ý ý ø 6 v ý  üö £ Ò – ¦ Æ È¦ È     † $qÈ £  ¼š–’|† $qÈ ¦ ý  ø ü ö Ò – Ò – © ¦ ¦  ý ø ü ö ¼ –’|  š ýü eŽö @ ' ý ø ü ö ¼ –’ï  š 6 v |5ü  €r ý ý ø ü G –Žö ù  › r w   6 dö  $¼š–’|† v ý øüö £ ý Å ø üö Æ P›–Ž† Æ Recall (2. ÷ G Åø ‰š÷ G Wü ø ý ø ü g –’ö Å š T and Therefore.1).2) (5. Brace-Gatarek-Musiela model (4.

6 Implementation of BGM £ ý Å ø ü Îý Å ø üö 5›–Žö —š$ŽïT Take ý ø ü ö  $Žï È Ð w gs à v u Plugging this into (5.4) (5. ‚ ‚ Ì úý Ì ø üö † i ý Ì v Å ø üö úý Å ø üö Î Ì ‹V¤›–ŽW¤¤y4š$Ž† eš–’¶¨|v £ ¥ È zg¥ s ‚ º v u  ú Vý¤Ì›ø–Žö†Wi¤ý¤Ì|vFŚø$Žö† (»I ü ü ý›ø–Žö† ü £ Ò – $qÈ ¦ âã ¥ i y   Ì ý ø ü ö  –Ž  › † %qÈ Ò – Obtain the initial forward LIBOR curve from market data.4) yields Note that (5.4’) (5. ÷ G Å ø ÷ G •©(¤©3ü ø ý ø ü g –’ö Å š T ýü eŽö @ V ýÅšø–’ö¶—ýÅ›ø–ŽöïTŒv ü Î ü ý Å ø ü ›–Žö Ó|v ÎÌ Å  ¥ ü ½ v ý Å ø ü 6ý Å ø ü Îý Å ø üö T v ý Å ø üö $š–’ö ù —›–Žö —š$ŽïŒ$š–’¶Î Æ Å › r 做–’ö r ÓÌ Tý Åøü Î ¹ Æ ý ø ü  –Žö  ý ø ü  –’ö ýÅ›ø–üŽö Ó|v ÎÌ ¥ $š–’ö ù 7ˆ¤|Fš$Žö ù 6 v ý Åøü 6 £ ýÌ v Åøü Tý Åøüö Π做–’¶¨Ì Å š ýü iŽö @ ' ý ø ü  –Žö Å › Îý Å ø üö T v —š$ŽïŒ%ü  ú ý V¤ Å ‚  Ì v Å ø ü 6ý Å ø ü Îý Å ø üö T v ý Å ø üö y4š–’ö ù —›–Žö —š$Žï|$š$Ž¶Î Æ Iš$Ž¶Î £ ý Å ø üö Æ ‚ ‚ Ì ¥ ¥ Then úý Åøü 6 i ý Ì v Åøü heš–’ö ù W´Dum›–Žö ù 6 ú ý ø ü ö V $Ž¶ Å š ÎÌ ¨uv úýü VeŽö @ ' v %ü  ý ¤ ‚ ‚ ‚ Ì Å   Ì v Åøü y4š$Žö ù 6 ‹eš–’ö ù W¤¤|Fš$Žö ù 6 V›–Žö ¨|v úý Åøü 6 i ý Ì v Åø ü ú ý Å ø ü ÎÌ v ý Åøü $ü ›–Žö Î Æ ¤›–Žö Î £ ý Åøü ¥ ¥ Æ Therefore.3) £ ý Åøü P›–Žö Î  Å š T Æ £ ý Åøü Æ ¤›–Žö Î Å Æ Æ .3) is equivalent to ß Å But 340 to be given by (5. Choose a forward LIBOR volatility function (usually nonrandom) ø ÷ © G Å øý Åø ÷ö (拁š¡¶Î 34.3’) (5.

the resulting Remark 34. we can for each ý ø ¤ ÷ Ì š Plugging the solution into (5.4) and (5. Brace-Gatarek-Musiela model With these functions. is usually taken to be nonrandom. We then solve (5.4’) become Å š Then and we continue recursively. 4. Returning to the HJM variables and .4’) to obtain (6. we obtain solve (5.2) that as ÷ ×Ì and so .3) we have  ˜ e’ö ý ü 6 “ 6  ý ô ø ü G–’ö ý ô ø ü G–’ö ý ¶ôG$Žö5ÓÌuv ø ü •  @'  ¥ v ýü i ôøü 5ü iX|G$Žö ù v $ü ý eüŸiWGø$ŽïT¢¶ôh–Ž5¨Ì ô ü ö ý ø ü ö • @ '  v ýÌ v ü i ôøü %ü ¤|5ŸËh–Žö ù ú ý ü VeŽö ‚ •ýü i ôø üö T ŠeŸrh–ŽïU£ •ýü i ôø üö T £ ý ôø üö ŠeŸrh–ŽïUI¶h–Ž5•   ü  ý ü iŸ i ôøüö |G–’¶Î Å Æ i ýü i ôøüö Æ ¤iŸ|G–’¶Î £ Pý ô ø ü G–’ö •  ýü i ôø ü Î £ ý ôø üö eŸyh–Žö aI¶G$Ž5• ô ü  qÌ ÅuzÌ 1 ý Å ø ü ›–Žö ù 6 o Å 1 ÷ ø ÷ G yA3ü ü Ì Ì o Å 1 ÷ ø ÷ G uA3ü Because LIBOR gives no rate information on time periods smaller than .3 From (5. we must also choose a partial bond volatility function for maturities less than from the current time variable . This is not as terrible as it first appears.1 BGM is a special case of HJM with HJM’s In BGM. ø ý ø ü g –’ö ù Å š CHAPTER 34.4) (equivalently. then Æ We saw before (eq. set ý ø ü ö  –’¶ ý ø ü  $Žö ù Å š ý ø ü ö  –Žï Å › T ý ø ü ö ù  –Žxñ 6 Å › 6 o Å 1 Ì ø ÷ G uzˆè©ü ø ý ø ü ö g –’¶ Å š Î ø ý ø ü ‹ –Žö ‚ õ Å › Î Ì Å 6 If we let and (5.1) 341 È Î Ï Ï . .2 (5.ø ý ø ü ö ‹ –Žï Å › ø ý Å v ü ø ü ‹|%h–Žö £ ý Å ø üö † çý Å ø üö Iš$Ž|—š$Ž¶Î ¸ × ÷ Ì eüŸi|h–ŽïŒ—¶h–Ž5—iXŒG–’ö T ý ô ø ü ö 6 ç ý ô ø ü ö • ý ü i ô ø ü ¼Ò ¼ Ì Ÿ ¨¤y4š–’ö ù 6 Æ —š$Ž¶Šš–’ö T ¼¼ ý Ì v Å ø ü çý Å ø üö Îý Å ø ü 6 7£  ý ø ü ö ù  $Žx‚ Å š Ì 6 i ý Ì v Å ø üö ù rI¤y4š$Žx6 ú ý ø ü ö e –’¶ Å š ÎÌ ¨yv ¥ ‚ £ ý Å ø üö Îý Å ø üö ¡š–’¶—›–ŽïT Remark 34. generated recursively by (5.4’) to obtain for Ì is ø¤Ìis o Remark 34. is random.3’). .4’)) is a stochastic partial differential equation because of the term.3’). (5.

the limit as 342 of (6.5): involving r • ü  ý ü eŸ i ôø üö 6 £ |G$Žïa¤ý ô ø ü G–’ö ¸ ÷  ×Ì Let  y  is a Brownian motion under Girsanov’s Theorem implies that This is a martingale. solutions @ ý ô ø ÷ ¶h3ö ý ô ø ü ¶G$Žö û ý ü heŽö û © £ ýü ¤iŽö .ø ô 5 1 ü 1 ”A÷ ø    ý   i ôø  W|hdö ù 6 … ¦ v ýü 5iŽö @   ¡ÿ þ £ ýü ˆe’ö    šý ô ö Ž¥  `õ  ý ÿþ ÿþ ô ø ÷ ö û ý ô G3©G¶Žö ý ô ø ô G¶’ö û ¢  © ¥  ý ô F¶Žö ¥ © ý ô ø ÷ ¶Gè3ö § ¨¦ û £ £ ¤ý ¢   ö ÿþ §¦  p   r ý ý G  i ôø  6 ŒËhdö ù dö … ¦ r @ '  Ihö i ý  ý  i ô ø  ŒËhdö ù 6 … ¦ i n v u gs £ ýü eŽö ú ý ô ø ü ö V¶h–Ž©û @ '  ý ü eX i ôøü |h–Žö ù 6 ý ü eŽö ý ô ø ü ö ¶h–Ž©û © i p£ ¥ © £ S ýü ëeŽö ý ô ø ü G–’ö v %ü  ý ô ø ü ¶G$Žö û ý ü ö heŽ † xi ‚ ý ü e’ö © û … © ý ý¶ôG$Ž5—eü |G–’ö r |1 ø ü ö • ý i ôøü T ý ô ø ü ö • ¶G$Ž5¨Ì 1 r —iX|G$Žö r XÌ •ýü i ôøü T ô ø ü G–’ö ý ô ø ü ¶G$Žö ý¶G$Ž5Ó|v ô ø ü ö • Ì ¥ r —eŸuG–’ö r qÌ •ýü i ôøü T – éè #é " … $  d! ˆ f " … qhÒ # "… f $  d† ê# …   d† ‚ @ V   úýü VeeŽö v ý i ôøü 5ü eü Ëh–Žö ù 6  ý Remark 34.7 Bond prices to this stochastic differential equation is given by From (2.1) has the term to this equation do not explode because ô ø ü G–’ö ¸ çý ôø üö —¶h–Ž5• so Therefore.1) is given by equation (2. and we can use it to switch to the forward measure … # †'& # "… $  d! Q `  ý ö    d † v u Ð w gs The solution 34. .6) we have .4 Although the term in (6.

1) (8. the floating rate interest the LIBOR at time .1) 343 .8 Forward LIBOR under more forward measure CHAPTER 34. At time . . Ì v urô ø ý ô g¶Gø ô ö Ž5 •Ì £ ý ÷ Óæ(x–ø Î ¨Ì   W|Gñdö r T … ¤iŽö r ‚ ý  i ô ø  £ ýü   ¦ r  @ V … ¤eŽ©’  ýWi|ôGøñdö r T …  ¤¼dö $– i ý  Ò ý  i ô ø  ö ¼rrhdïT £ ýüö     ¦   ¦ r º v u  @ ' … mgs `»P ý¼WiËôhødö r T …  ¤dö %– i ý  Ò ý  i ô ø  ö rŒGñdïT     ¦   ¦ » I  ý ¼ ý ô ø ü ö ¶h–Ž5 • a£ • £ a5ý ô ø ô G¶’ö  i ôø  r|Gñdö r T   ¦ r  Ihö $– i ý  Ò   @ V ý ¼  i ôø ö WrGñdïT º v u mgs   ¦ v u gs ý ô ø ÷ ö ¶h35 p   rŒGñdö r T ý  i ô ø  … ¦ r  Ihö $– i ý  Ò   @ V ý ¼  i ôø ö WrGñdïT n … ¦ ý ô ø ÷ ö ¶h35 • £ ý ôø üö aI¶G$Ž5• so ú ý ü ViŽö @ ' v $ü  ý ¤ @ Ò $– V ø ý ü ‹e’ö   ‚ Ì v ü i ôøü u%X|h–Žö ù 6 ý ô ø ü ö ¶G$Ž5 eü ý ý¶ôGø$Žö5•—eü rh–ŽïU£ ü ý i ôø üö T  • i ôø üö T £ ý ôø üö — rh–ŽïUI¶G$Ž5• We assume that T and mean ý ü iŽö r r ‚ i • and Case I: is normal with variance 34. the value of the caplet at time is .2) (8. Then .1) we have is nonrandom.9 Pricing an interest rate caplet From (6. We determine its value at times . ý Ì v ôø ü û –ý Ä i ý ôø ôö •ö Ì D|ŒG–’ö R—‹r$¶h¶Ž5da£ ¼¼ ¤yWŽö ý Ì v ô ˜ ýiŽö£ ¼¼ – ‹r$¶h¶Ž5da£ ü ý Ä i ý ô ø ô ö • ö Ì © ý ü eŽö “ ’þ © ¼¼ ý Ì v ô ¤|rŽö ˜  ¼¼ – ‹r$¶h¶Ž5dÌ ý Ä i ý ô ø ô ö • ö £ ýü Ò ¤iŽö %– © ý ü e’ö “ ’þ   © ý ü ö iŽ£ ì Ìyvr”5aô ô 1 ü 1 – ‹W$¶h¶Ž5dÌ ý Ä i ý ô ø ô ö • ö ô ö Ž¶ Ì v yrô ÌyvŒô51ü”1A÷ ÄÌ k ¨Uˆý ô ô ø ô G¶’ö • ¨Ì Ä ¨Ì Consider a floating rate interest payment settled in arrears. Thus. Brace-Gatarek-Musiela model 34. A caplet protects its owner by payment due is requiring him to pay only the cap if . (9.

    5ÿ þ .1) and (8. we have û e T Ì £ ýü Ò 7¤eŽö $– ì ’   5ÿ þ ý ü ö iŽ© ’ û ¤ Ì a£ û ¤ Ì a£ £ ýü Ò ¤iŽö $–   ì ý ¤ Ì v ô |rŽö ý ü eŽö © © “ ’þ Then (9.ø ü –X i uô  T £ ýüö U¤iŽ‚ Ì i ô 1 ü 1 Xry535A÷ ø ý ý ô ø ü ö gG¶G$Ž5 1 D ö ý r r S S Ä ~i†{IeŽ‚ Q } ýüö Ä i†5iŽ‚ Q ~ }{ ý üö  ýiŽö‚  i ü 3ÄWi e’ö ‚  v ý ü  ù á7£ ù ù ¥ v u gs ¥ w € – ‹Wx›eŽ©qB ý Ä i E ý ü ö ’ ù Ò dö $– £ ý ùö Pzd1b   x’ þ Set ¼¼ v u gs  ˜ eŽ£ ¼¼ – ‹Wx›iŽ©qB ý ü ö ý Ä i E ý ü ö ’   Ð ’ qB ý ô ø ü ö ¶G$Ž5 ö Ò • Ò ª %– ý Ì v ô ø ü ö û Ì ¤yrh–Ž©D7¤eŽö £ ý ü $– “   x’ þ   ì  Ò %– ý ü ö eŽ© ý ü ö e’ ø ý ü ö  eŽ© E ø   £ ýü ˆiŽö r ‚ v u gs ý ô ø ü ö ¶G$Ž5 ã ¼ ÷÷ ¼¼ ý Ä i ý ô ý ü ö ·÷÷â eŽ£ ¼ – ‹W¡¶Gø Ò $–  #  $ö à ý¤Ìyõh3©Á  G¤yrŽö v ô ø ÷ ö ›û ý Ì v ô ô Žö5ƒö • Ì v ôø Ì v ôö uÔh8urŽ©û ý ¤ ¼¼ Ò ˜ ýeŽö£ ¼¼ – ‹W5¶h¶Ž5dö $– ü ý Ä i ý ô ø ô ö • ý Ì v ô ø ü ¤urG$Žö “   ’þ óò·ñiï ðî À ôàô à ¤|rG$Žö û À  ý Ì v ›ôÁ ø ü ý Ì v ô ø ü © ßôô ’ þ D|Wh3ö heŽö ¤urG$Žö ý Ì v ô ø ÷ û ý ü © ¼¼ ˜ eŽ¥ x—‹Œ$¶Gø ý ü ö ¼¼ – ý Ä i ý ô ô ö Ž5  ö ø ý ‹¤ Ì v ôö  |rŽ£(õ where is normal under with variance Furthermore.2) we have  ý ¤ Ì v ô yŒŽö . í § A¨¦ ô 5 £ ¤ý ¢ Case II: Recall that 1 ü 1 35A÷ We have ö Ò %– where 344 and mean ý ü eŽö r r ‚ i ý   i ôø  Œ|Ghö r T … • £ ý ôø ôö a5¶h¶Ž5• and (9.2) is called the Black caplet formula.2) . is independent of . • d Ì ý¤ÌyWG¡ö GiŽö v ôø ÷ ûýü  © eŽö £ ýü Ì v ôø üö yrh–Ž©û í ý ¤ ¢  ø ÿþ From (8. • b Ì v ôøü d ¡ urG$Žö In the case of constant .

e.10 Pricing an interest rate cap CHAPTER 34.e.1) 345 b . If we (11. is a deterministic function of its second argument.11 Calibration of BGM The value at time of the cap is the value of all remaining caplets. i.. we can “back out” is constant on each of the intervals has time-zero value .ø ý gy hø ô ú $  ˆ $Ž¶Xqxgý  ú ôö ø    ø ô Gø ô ö ø ’Ègý ô ø ÷ G3ö  YÉ  É dr T î ý z ö ¢     ¦ ø ¼  qq ø    ¡  É  ý z ö É dr T ÿ X  ¦ ø  i FÉ  ý z ö É dr T î   ¦ ø ý ÷ ‹x3ö ø ý ÷ ‹x3ö Ò $–    9É  ý z ö É dr T £ R  ì  ì ý   i ô Œ|Žö r T  ý Åö T £ ý Åøü dïU¡š–’ö T  Let us suppose  ¡ ý ¼  i ôø  W|Ghö r T   ¦ b   ø ý ý ô ø ÷ ö gG¶Gè35 ñ ¤ ö ý • b Ì v ôø ÷ö û Ì £ ý ÷ Ò d ¡ |rh3©D7Px¡ö $– at time ì Î ýü eŽö þ  þ ý ¨  h… ¤ ü — £ ýü eŽö ü ¥ ¥ Ì û £ ú ô ø  øÌ s X$aá$¼qXx8„a£ r ô 8 ø Ì a£  ô ø ÷ Rl £ ô A cap is a series of payments 34. . i. we can “back out” the squared volatility É  ý z ö É dr T Ð   ¡  øý ÷ Ò XqX‹x3ö $– î   ì Ò $– ì ý ÷ x3ö ÿ X  Ò $– T where   £zÉ ýzdö r T ÿX  ¼É dö r T É ø ý É î ¦ ÿ ¦ o o X   qX  ô ô ú $ô o  In this case. we may assume that T .. Brace-Gatarek-Musiela model Let at time  i ûø  `šXqø ø ÷ l ì — £ – 2¤ø  – ô ì — –ýÄ i —‹rIý ô Gø — The interest rate caplet on Ì v yrô ý ô ø ÷ ö ¶h3¶ Ä ô ö Ž5  ö • d Ì Then depends on   ¦   ¦ If we know the caplet price know caplet prices where (defined in the last section) depends on 34.

we can “back out” for all . The long rate is 34. Then be a large fixed positive integer.12 Long rates û The long rate is determined by long maturity bond prices. so where the last equality follows from (4. and We can now solve (or simulate) to get ø ÷ © or equivalently.     ì and then differen- Í Å hr T (see (2.346 and choose these constants to make the above integrals have the values implied by the caplet prices. using the recursive procedure outlined at the start of Section 34.6)).1). ý ø ü  –Žö ù Å › Ì Ì $û 6 . ø ÷ © 34. Let that is 20 or 30 years. To implement BGM.6. it is reasonable to set Since is small (say year). we need both .13 Pricing a swap ÷ s Ì û X$|v ô q £ ú ô ø  øÌ s A$XXq|¤iyv ô q £ r ô ø X¤ Ì |v ô ‘ £  ô G ô Let be given. and set Å 7v ü Now is the volatility at time of a zero coupon bond maturing at time . and É  ý z ö úý Ì V¤—ý É dr T ¥  — ‚ Ÿ ø’eDŠý ú ý Ì i¿dD$Ž¶¨|v –ö ø üö Î Ì £ ¥ ¥ ¤ ú Ò  ˆ — zgs  — º v u Ÿ  » P ý¼š$üŽ† ø ö £ # җ ¦ ú¤ º v u ¿gs  »I ›ø–Žö† Ò ý ü £ ú ¦ Ð i –ö øü zd¤–Žö  XÌ  XÌ o Å 1 ÷ ø ÷ G yAX詏ü o  — ‚$~i}†{ Ÿ $û ~ } Ì i†{ %û Ì ý Ì û ø ü ¤$š–’ö ΨÌyv £ ú¤ ¥ ¥ Ç ¥ Å 1 ÷ ø ÷ G uA3ü G ôø ÷ G UGsü G Åø ÷ G `›©3ü ÷ © ¥ G (Å ÷sG(ÅP‹hö T ø ý Å Å dr T ÷ © ø ‹ý ø ý ø ü ‹ –Žö ø ÷ l  $Žö ù ý ø ü G aô Ì ø ý ø ü g –’ö ù ý  ö Å › ô ø ü G–’ö o £ Å I š Å š Å 1 yŽ÷ ü Î • ý ¤ £ Ì ûøü $š$Žö 6 ý ÷ x¡ö £IýdïT Åö Ò $– ¥ 6 Ç If we know caplet prices tiate to discover and ý  ö for all .

T  which makes the timeü ú Vý û ‚ Ì ü ü v ôøü ú ôhø–Žö©û5vøXqU5ý  G$Žö i ôø üö ¤ h–Ž©û ý  G$Žö ô ø ü ú $ û X ý £ ýü eŽö ÿ X  § û ý G‹ ÄÌ ¨|v ¥ ö © ý i ¤  ˆ $G$Žö ú ôøü ÿX  ý ú G–ü’ö ûXi¤ý ú ôh–Žö©‹ÓÌWi¨Xq|i¤ý r ôG–’ö û‹¨Wi¤ý  h–Ž©‹Ä¨W¤ý h–Ž©l£ ô ø ø ü û Ä  ø ü ÄÌ ô ø ü ö û Ì i ô ø ü ö û r ûsvšqXUv5ý  ô ü hø–Žö©Gý‹¨Ì|v û Ä ö i¤ý  Gø$Žö û55ý  G$Žö G‹Ä¨|v ©i¤ý h–Ž©l£ ô ü v ôøü ûý Ì ö ô ø ü ö û ¥ ¥ —  – — h–Žö©G‹¨Ìyv öqPý — G–’ö û ‚ Ÿ £ ú eý ô ø ü û ý Ä i ôøü ¥  ˆ ¤ú —  – — ’ö ¼¼ ý ô — Ÿ ˜ iüŽ£ ¨ý‹Œi5x$ø ’¶dÌ ý ö ¼¼ Ä ý ÷ ô ö Î ö © “ ’þ  ˆ ¤ ú for maturity ô ö ’¶  ö ý — —  ý  – h–Ž©G‹¨yv ö i ô ø ü ö û ý Ä Ì ¼ ¥ q– ¤ý  — ’ö ô  –— ¼¼ ý ý ô ø ü ö û ý Ä Ì h–Ž©G‹¨|v ö ©i ˜ e’ ¼ ý ü ö ¥ ý ü iŽö © ¼ # î ¼ ¥¦Ó   " þ–¨—    þ  ÷ã À ý  – — hø — Ž©Gý — Žö ô ô ô ö û ô  – — ôhø–Žö©ûGý‹Ä¨Ìyv öqi ÷ ü ¼¼ à — ô ¼¼ ý ›Á Žö ü ·÷÷â ýeŽö¥ ¼ ˜ ý Žö£ ¼ ý — ’ö © “ ’ þ ¥ ô ý ü iŽö © ©  – — Gø — Ž©û Q ý  – — ’ö ¼¼ S ý ô ô ö ô ˜ ýe’ö ¼¼ ¨ÌWi i ü Ä ¥ ý ü iŽö  ©– ¼¼ ¥ — ý ˜ iŽ£ ¼¼ ‹ŒPý –ø Ž¶h|Ì ý ü ö ý Ä i ÷ ô ö Î ö  –— ý  m˜ ¥ ¼¼ —  ˜ýüö ý Ä i ý ÷ mhe’¥ ¼¼ ‹WIx$ø Î d Ì ý  –— ý ü iŽö ô Žö © ¥ ¥ The swap is the series of payments CHAPTER 34. which depends on the term structure model and requires estimation of . the value of the swap is is the value of Ä ô § ý ü eŽö © ü ô ø ü G$Žö û Ò £ © “ ’þ £ ôàô £ © ßôô ’þ © — “ ’þ ô ’ö £ “ ’þ — ý ü eŽö © © ô ö Ž¶ Î 347 . the swap formula is generic.ý  G$Žö ô ø ü ú $ In contrast to the cap formula. Brace-Gatarek-Musiela model at time  i ûø  R›qXø ø ÷ Ò — £ – á›ø  – ô The forward swap rate at time value of the swap equal to zero: ý ü eŽö i ø © ý ‹ — Ÿ “ ”’ þ  ˆ ¤ ú Ä i r5ý ÷ –ø The value of the swap at time is ô Gø — ô Žö û “ ¥ ô Žö Ì £ ý ÷ Px$ø û ý ¥  –— ô Gø — — ô ö Ž¶ | ö Î h Ì For ô ” 1 ü 1 5A÷ Now ¥ £ ý ÷ Px$ø We compute — ô ö Ž¶ ÎÌ ¨yv ¥ so .

348 .

although the concept dates back to 1934 work of L´ vy. The binomial asset pricing model was developed by Cox. In this stage one views a random variable as a function from a sample space to the set of real numbers . Furthermore. A well-written book which contains all these things is DeGroot (1986). This point of view handles both discrete and continuous random variables within the same unifying framework. A more detailed book is Chung (1968). 349  ¢  6 6 ý ¢ ö ÿþ . measurable with respect to the conditioning -algebra. The first complete account of martingale theory is Doob (1953). These can be used to compute expectations and variances. and even conditional expectations. Probability theory is usually learned in two stages. e ¨ ©þ ¨ 35. A conditional expectation is itself a random variable. Each set in has a probability . In the first stage. Dothan (1990) and Ritchken (1987). and then binomial versions are developed for purposes of implementation.1 Probability theory and martingales. including Cox & Rubinstein (1985). Accounts of this model can be found in several places. The term martingale was apparently first used by Ville (1939). The second stage of probability theory is measure theoretic. Many models are first developed and understood in continuous time. and the collection of all events forms a -algebra .2 Binomial asset pricing model. Ross & Rubinstein (1979). one learns how transformations of continuous random variables cause changes in their densities. The measure-theoretic view of probability theory was begun by Kolmogorov (1933). Certain subsets of are called events. one learns that a discrete random variable has a probability mass function and a continuous random variable has a density.Chapter 35 Notes and References 35. 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). This point of view is indispensible for treating the rather complicated conditional expectations which arise in martingale theory.

a much weaker condition than the claim that stock prices follow a geometric Brownian motion. According to this hypothesis. The development in this course is a summary of that found in Karatzas & Shreve (1991). It was introduced to o finance by Merton (1969). 1924). 1948) discove ered many of the nonintuitive properties of Brownian motion. Some mathematics texts on stochastic calculus are Øksendal (1995). 1973) presents the argument that geometric Brownian motion is a good model for stock prices. A provocative article on the source of . The first mathematically rigorous construction of Brownian motion was carried out by Wiener (1923. This is often confused with the efficient market hypothesis. past stock prices may be useful to estimate the parameters of the distribution of future returns. and Jarrow (1988). A mathematical construction of this integral.5 Stochastic calculus and financial markets. L´ vy (1939. Alexander (1961) and Fama (1965). but they do not provide information which permits an investor to outperform the market. is Musiela & Rutkowski (1996). now in preprint form. which asserts that all information which can be learned from technical analysis of stock prices is already reflected in those prices. Dothan (1990) and Duffie (1992). Some empirical studies supporting the efficient market hypothesis are Kendall (1953). This motion is now known to be caused by the buffeting of the pollen by water molecules. The mathematical formulation of the efficient market hypothesis is that there is a probability measure under which all discounted stock prices are martingales. Stochastic calculus begins with Itˆ (1944). Chung & Williams (1983). A criticism of the efficient market hypothesis is provided by LeRoy (1989). In 1828 Robert Brown observed irregular movement of pollen suspended in water. 35. with a minimum of fuss. Brownian motion and its properties are presented in a numerous texts. Boness (1964). Bachelier (1900) used Brownian motion (not geometric Brownian motion) as a model of stock prices. Sprenkle (1961). include sections on Itˆ ’s o integral and formula. even though Brownian motion can take negative values. Samuelson (1965. The integral with respect to Brownian motion was developed by Itˆ (1944). The last of these papers discusses other distributions which fit stock prices better than geometric Brownian motion.4 Stochastic integrals. is given by Øksendal (1995).3 Brownian motion.350 35. including (in order of increasing o mathematical difficulty) Hull (1993). Osborne (1959). An excellent reference for practitioners. Many finance books. Protter (1990) and Karatzas & Shreve (1991). including Billingsley (1986). as explained by Einstein (1905). 35. Some other books on dynamic models in finance are Cox & Rubinstein (1985). The quadratic variation of martingales was introduced by Fisk (1966) and developed into the form used in this course by Kunita & Watanabe (1967). Huang & Litzenberger (1988). Ingersoll (1987).

which makes good reading even today. An important companion paper is Merton (1973). and risk-neutral measures. the martingale representation theorem. A numerical treatment of the partial differential equations arising in finance is contained in Wilmott. The other books cited previously. so that is a probability measure. see Karatzas & Shreve (1991). due to F¨ llmer & Schweizer (1993). 1995) and also Duffie (1992). Notes and References stock price movements is Black (1986). The form of the martingale representation theorem presented here is from Kunita & Watanabe (1967). connecting stochastic differential equations to partial differential equations. The constant elasticity of variance model for option pricing appears in Cox & Ross (1976). ÿþ  ¥ £ ý ô ¤¶Žö í ’þ  . the Black-Scholes option hedging formula seems not to be very sensitive to deficiencies in the model. and Karatzas & Shreve (1991). including discussions of the Kolmogorov forward and backward equations. The application of the Girsanov Theorem and the martingale representation theorem to risk-neutral pricing is due to Harrison & Pliska (1981). It can also be found in Karatzas & Shreve (1991). page 182. This methodology frees the Brownian-motion driven model from the assumption of constant interest rate and volatility. has o the geometric Ornstein-Uhlenbeck process as a special case. Dewynne and Howison (1993. 35. Markov processes which are solutions to stochastic differential equations are called diffusion processes. (This and many other papers by Merton are collected in Merton (1990). all treat this subject. Øksendal (1995). 35. the Brownian-motion driven model is mathematically the most general possible for asset prices without jumps. Another alternative model for the stock price underlying options. is due to Feyman (1948) and Kac (1951).7 Girsanov’s theorem. or even through dependence on the paths of other assets. o is similar to that originally given by Black & Scholes (1973). Protter (1990). Girsanov’s Theorem in the generality stated here is due to Girsanov (1960).CHAPTER 35. A good introduction to this topic. When both the interest rate and volatility of an asset are allowed to be stochastic. The Feynman-Kac Theorem. these parameters can be random through dependence on the path of the underlying asset. Chung & Williams (1983). using only Itˆ ’s formula. although the result for constant was established much earlier by Cameron & Martin (1944). The theorem requires a technical condition to ensure that .6 Markov processes.) Even though geometric Brownian motion is a less than perfect model for stock prices. page 198. is Chapter 15 of Karlin & Taylor (1981). 351 The first derivation of the Black-Scholes formula given in this course. Kloeden & Platen (1992) is a thorough study of the numerical solution of stochastic differential equations.

Schweizer (1992a. and Geman & Yor (1993) obtain the Laplace transform of the price. most work has been devoted to approximations. Some works on hedging and/or optimization in models which allow for jumps are Aase (1993). Madan & Milne (1991). but in continuous-time models. Delbaen (1992). Lakner (1993). Rogers & Shi (1995) provide a lower bound. Delbaen & Schachermayer (1994a.b). Brennan & Schwartz (1977) (see Jaillet et al. This is true in discrete-time models. Elliott & Kopp (1990). by Cox. Merton (1976). pages 196–197. Jacka (1992). that the absence of arbitrage implies the existence of a risk-neutral measure.1991) and Xue (1992).1992). Delbaen (1992). Bates (1988. Bouaziz. see Myneni (1992) for a survey and additional references. 35. Sosin & Gatto (1979). We do not provide an explicit pricing formula here. Jarrow & Madan (1991a). Naik & Lee (1990). 1995). a slightly stronger condition is needed to guarantee existence of a risk-neutral measure. e Because it is difficult to obtain explicit formulas for the prices of Asian options. Mercurio & Runggaldier (1993). including Stricker (1990). M¨ ller (1989) and Taqqu & Willinger (1987). Lookback options have been studied by Goldman. The Fundamental Theorem of Asset Pricing. For the continuous-time case. u 35.352 When asset processes have jumps. In addition to the fundamental papers of Harrison & Kreps (1979).b). Back (1991). Goldman. adjusted to account for drift.9 American options. risk-free hedging is generally not possible. Jones (1984). which leads to a justification of the Brennan-Schwartz algorithm). Ross & Rubinstein (1979) (see Lamberton (1993) for the convergence of the associated binomial and/or finite difference schemes) .. as stated here.e.8 Exotic options. Sosin & Shepp (1979) and Conz´ & Viswanathan (1991). The reflection principle. A general arbitrage-based theory for the pricing of American contingent claims and options begins with the articles of Bensoussan (1984) and Karatzas (1988). Approximation and/or numerical solutions for the American option problem have been proposed by several authors.c). Explicit formulas for the prices of barrier options have been obtained by Rubinstein & Reiner (1991) and Kunitomo & Ikeda (1992). (1990) for a treatment of the American option optimal stopping problem via variational inequalities. 1983). including Black (1975). can be found in Harrison & Pliska (1981. The perpetual American put problem was solved by McKean (1965). Jarrow & Madan (1991b. is taken from Karatzas & Shreve (1991). Beinert & Trautman (1991). Bryis & Crouhy (1994) provide an approximate pricing formula. some other works on the relationship between market completeness and uniqueness of the risk-neutral measure are Artzner & Heath (1990). although the partial differential equation given here by the Feynman-Kac Theorem characterizes the exact price. and Harrison & Pliska (1981. and Fritelli & Lakner (1994. It is tempting to believe the converse of Part I. Madan & Seneta (1990). 1983). Shirakawa (1990. i. results have been obtained by many authors.

35. The Cox-Ingersoll-Ross model is presented in (1985a. Brace & Musiela (1994a. and Carr & Faguet (1994).b).12 Change of num´ raire. Broadie & Detemple (1994). which also considers options on futures. Other surveys of term structure models are Duffie & Kan (1994) and Vetzal (1994). and Longstaff & Schwartz (1992a. The presentations of these given models here is taken from Rogers (1995). A summary of this model appears as Reed (1995). . Black (1976)) and even more recently understood. Bunch & Johnson (1992). A partial list of other term structure models is Black. Carverhill & Pang (1995) discuss implementation. and a discrete-time version is provided by Heath. The prices for differential swaps have been worked out by Jamshidian (1993a. Cox. The continuous-time Heath-Jarrow-Morton model appears in Heath. et al.b). Omberg (1987). equilibrium setting.CHAPTER 35. Jamshidian (1990). Duffie & Kan (1993). Foreign exchange options were priced by Biger & Hull (1983) and Garman & Kohlhagen (1983).b). Barone-Adesi & Whalley (1987). whereas Richard & Sundaresan (1981) study these claims in a continuous-time. The Brace-Gatarek-Musiela variation of the HJM model is taken from Brace. The Merton option pricing formula appears in Merton (1973). 35. Similar ideas were used by by Jamshidian (1989). 35. Related works on term structure models and swaps are Flesaker & Hughston (1995) and Jamshidian (1996).13 Foreign exchange models. Jarrow & Morton (1990). and to Chapte 7 of Duffie (1992). 1982) (but see Hogan (1993) for discussion of a problem with this model). one may consult Duffie (1989). The distinction between futures contracts and daily resettled forward contracts has only recently been recognized (see Margrabe (1976). For additional reading on forward and futures contracts. 35. Ho & Lee (1986). Notes and References 353 and by Parkinson (1977).b). MacMillan (1986). 1993b) and Brace & Musiela (1994a). Ingersoll & Ross (1981) and Jarrow & Oldfield (1981) provide a discrete-time arbitrage-based analysis of the relationship between forwards and futures.10 Forward and futures contracts.11 Term structure models. Geske & Johnson (1984). e This material in this course is taken from Geman. (1995). Our presentation of this material is similar to that of Duffie & Stanton (1992). Johnson (1983). El Karoui and Rochet (1995). The Hull & White (1990) model is a generalization of the constant-coefficent Vasicek (1977) model. Brennan & Schwartz (1979. Barone-Adesi & Elliott (1991). Jarrow & Morton (1992). Implementations of the model appear in Hull & White (1994a. Derman & Toy (1990).

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), 65–84. (1961) Alexander, S. S., Price movements in speculative markets: trends or random walks, Industrial Management Review 2, 7–26. Reprinted in Cootner (1964), 199–218. (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, 21–86. Reprinted e e in Cootner (1964). (1991) Back, K., Asset pricing for general processes, J. Math. Econonics 20, 371–395. (1991) Barone-Adesi, G. & Elliott, R., Approximations for the values of American options, Stoch. Anal. Appl. 9, 115–131. (1987) Barone-Adesi, G. & Whalley, R., Efficient analytic approximation of American option values, J. Finance 42, 301–320. (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, 269–280. (1984) Bensoussan, A., On the theory of option pricing, Acta Appl. Math. 2, 139–158. (1983) Biger, H. & Hull, J., The valuation of currency options, Financial Management, Spring, 24–28. (1986) Billingsley, P., Probability and Measure, 2nd edition, Wiley, New York. (1975) Black, F., Fact and fantasy in the use of options, Financial Analyst’s J., 31, 36–41, 61–72.

CHAPTER 35. Notes and References
(1976) Black, F., The pricing of commodity contracts, J. Financial Economics 3. (1986) Black, F., Noise, J. Finance 41, 529–543.

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, 33–39. (1973) Black, F. & Scholes, M., The pricing of options and corporate liabilities, J. Political Economy 81, 637–659. (1964) Boness, Some evidence of the profitability of trading put and call options. In Cootner (1964), 475–496. (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, 259–283. (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, 449–462. (1979) Brennan, M. J. & Schwartz, E. S., A continuous-time approach to the pricing of bonds, J. Banking Finance 3, 133–155. (1982) Brennan, M. J. & Schwartz, E. S., An equilibrium model of bond pricing and a test of market efficiency, J. Financial Quantitative Analysis 17, 301–329. (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 efficient valuation method for American puts, using a modified Geske-Johnson approach, J. Finance 47, 809–816. (1944) Cameron, R. H. & Martin, W. T., Transformation of Wiener integrals under translations,

356 Ann. Math. 45, 386–396. (1994) Carr, P. & Faguet, D., Fast accurate valuation of American options, preprint, Johnson School of Business, Cornell University. (1995) Carverhill, A. & Pang, K., Efficient and flexible bond option valuation in the Heath, Jarrow, and Morton framework, J. Fixed Income, September, 70–77. (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, 1893–1907. (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, 321–346. (1985a) Cox, J. C., Ingersoll, J. E. & Ross, S., An intertemporal general equilibrium model of asset prices, Econometrica 53, 363–384. (1985b) Cox, J. C., Ingersoll, J. E. & Ross, S., A theory of the term structure of interest rates, Econometrica 53, 373–384. (1976) Cox, J. C. & Ross, S. A., The valuation of options for alternative stochastic processes, J. Fin. Economics 3, 145–166. (1979) Cox, J. C., Ross, S. A. & Rubinstein, M., Option pricing: a simplified approach, J. Financial Economics 7, 229–263. (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), 107–130.

CHAPTER 35. Notes and References

357

(1994a) Delbaen, F. & Schachermayer, W., A general version of the fundamental theorem of asset pricing, Mathematische Annalen 300, 463–520. (1994b) Delbaen, F. & Schachermayer, W., Arbitrage and free-lunch with bounded risk, for unbounded continuous processes, Math. Finance 4(4), 343–348. (1953) Doob, J., Stochastic Processes, Wiley, New York. (1990) Dothan, M. U., Prices in Financial Markets, Oxford University Press, New York. (1989) Duffie, D., Futures Markets, Prentice-Hall, Englewood Cliffs, NJ. (1992) Duffie, D., Dynamic Asset Pricing Theory, Princeton University Press, Princeton. (1993) Duffie, D. & Kan, R., A yield-factor model of interest rates, preprint. (1994) Duffie, D. & Kan, R., Multi-factor term structure models, Phil. Transactions Royal Society London A 347, 577–586. (1992) Duffie, D. & Stanton, R., Pricing continuously resettled contingent claims, J. Economic Dynamics and Control, 16, 561–573. (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, 429–444. (1965) Fama, E. The behavior of stock-market prices, J. Business 38, 34–104. (1948) Feynman, R., Space-time approach to nonrelativistic quantum mechanics, Rev. Mod. Phys. 20, 367–387. (1966) Fisk, D. L., Sample quadratic variation of continuous, second-order martingales, Z. Wahrscheinlichkeitstheorie verw. Gebiete 6, 273–278. (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, 1–23.

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 finance market model: the fundamental theorem of asset pricing, IMA Vol 65: Mathematical Finance, 89–92, (M. Davis, D. Duffie, W. Fleming & S. Shreve, eds.), Springer-Verlag, New York. (1983) Garman, M. & Kohlhagen, S., Foreign currency option values, J. International Money Finance 2, 231–237. (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, 443–458. (1993) Geman, H. & Yor, M., Bessel processes, Asian options, and perpetuities, Mathematical Finance 3, 349–375. (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, 285–301. (1979) Goldman, B., Sosin, H. & Gatto, M., Path dependent options: “Buy at the low, sell at the high,” J. Finance 34, 69–84. (1979) Goldman, B., Sosin, H. & Shepp, L. A., On contingent claims that insure ex-post optimal stock market timing, J. Finance 34, 401–414. (1979) Harrison, J. M. & Kreps, D. M., Martingales and arbitrage in multiperiod securities markets, J. Economic Theory 20, 381–408. (1981) Harrison, J. M. & Pliska, S. R., Martingales and stochastic integrals in the theory of continuous trading, Stochastic Processes Applications 11, 215–260. (1983) Harrison, J. M. & Pliska, S. R., A stochastic calculus model of continuous trading: complete markets, Stochastic Processes Applications 15, 313–316. (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, 419–440. (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, 77–105.

359

(1993) Hogan, M., Problems in certain two-factor term structure models, Annals Applied Probability 3, 576–581. (1986) Ho, T. S. Y. & Lee, S.-B., Term structure movements and pricing interest rate contingent claims, J. Finance 41, 1011–1029. (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, 573–592. (1994a) Hull, J. & White, A., Numerical procedures for implementing term structure models I: single-factor models, J. Derivatives, Fall, 7–16. (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 & Littlefield. (1944) Itˆ , K., Stochastic integral, Proc. Imperial Acad. Tokyo 20, 519–524. o (1992) Jacka, S., A martingale representation result and an application to incomplete financial markets, Math. Finance 2, 239–250. (1990) Jaillet, P., Lamberton, D. & Lapeyre, B., Variational inequaltiies and the pricing of American options, Acta Appl. Math. 21, 263–289. (1989) Jamshidian, F., An exact bond option pricing formula, J. Finance 44, 205–209. (1990) Jamshidian, F., The preference-free determination of bond and option prices from the spot interest rate, Advances Futures & Options Research 4, 51–67. (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, 59–60. (1988) Jarrow, R., Finance Theory, Prentice Hall, Englewood Cliffs, NJ. (1991a) Jarrow, R. & Madan, D., A characterization of complete security markets on a Brownian filtration, Math. Finance 1(3), 31–43. (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. & Oldfield, G., Forward contracts and futures contracts, J. Financial Economics 9, 373–382. (1983) Johnson, H., An analytic approximation for the American put price, J. Financial Quant. Analysis 18, 141–148. (1984) Jones, E. P., Option arbitrage and strategy with large price changes, J. Financial Economics 13, 91–113. (1951) Kac, M., On some connections between probability theory and differential and integral equations, Proc. 2nd Berkeley Symposium on Math. Stat. & Probability, 189–215. (1988) Karatzas, I., On the pricing of American options, App. Math. Optimization 17, 37–60. (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, 11–25. Reprinted in Cootner (1964), 85–99. (1992) Kloeden, P. E. & Platen, E., Numerical Solution of Stochastic Differential Equations, SpringerVerlag, New York.

N. E. Option pricing with V. G.. A theory of forward and futures prices. (1967) Kunita. (1965) McKean. Economic Literature 27. model for share returns. F. (1986) MacMillan. Math. 16–23. & Milne. & Schwartz. E.. The V. 39–56. Jr. A. Grundbegriffe der Wahrscheinlichkeitsrechnung. P. New York. D.. 1950. English translation: Foundations of Probability Theory. & Watanabe.. preprint. 275–298. On square-integrable martingales. J. Finance 3. G.. F. 119–139. e e (1948) L´ vy. H. Samuelson (1965): A free boundary problem for the heat equation arising from a problem in mathematical economics. Analytic approximation for the American put option. J. A. Nagoya Math. Math.. Martingale measures for a class of right-continuous processes. Chelsea. (1992) Kunitomo. (1990) Madan. Paris.. (1939) L´ vy. P.. S.. Interest rate volatility and the term structure: a twofactor general equilibrium model. L. Gauthier-Villars. W. A two-factor interest rate model and contingent claim valuation. A. Mathematical Finance 2. 43–53. Business 63.. University of Pennsylvania. Math. Compositio Math. & Seneta. 7. Convergence of the critical price in the approximation of American options. (1991) Madan.CHAPTER 35. Wharton School of Business. Processus Stochastiques et Mouvement Brownian. martingale components. P. 1583– 1621. D.. 1259–1282. Futures Options Research 1. (1993) Lakner. J. J.. Pricing options with curved boundaries. E. 209–245. Notes and References 361 (1933) Kolmogorov. 283–339. F. N. D. 179–190. H. A. P. e (1992a) Longstaff. 511–524. Finance 3. & Ikeda. M. Sur certains processus stochastiques homog´ nes. F.. (1976) Margrabe. Efficient capital markets and martingales. & Schwartz.. S.. Journal 30. Industrial Management Review . Finance 1(4). Appendix to P.. Ergebnisse Mathematik 2. S. (1993) Lamberton. S. (1989) LeRoy. Adv. Finance 47. (1992b) Longstaff. Fixed Income 3.

General equilibrium pricing of options on the market portfolio with discontinuous returns. 21–36. & Lee. W. u Economic Letters 31 37–41.. B. Arbitrage Pricing of Derivative Securities. Basil Blackwell. Lifetime portfolio selection under uncertainty: the continuous-time case. N. Fin. (1987) Omberg. Finance 3. 191–200. (1990) Merton. to appear. Economics 3. Rev. New York. & Rutkowski.. Option pricing for jump diffusions: approximations and their interpretation. M. F. Option pricing when underlying stock returns are discontinuous. C. R. (1993) Mercurio. & Runggaldier.. (1990) Naik. M. 141– 183. R. 117–142. Financial Studies 3. 493–521. 247–257. (1990) Protter. Risk 8(8). M. Stochastic Integration and Differential Equations. (1959) Osborne.. Springer-Verlag. 125–144. 32–39. Operations Research 7... Option pricing: the American put...  qX (1995) Reed. Math. (1969) Merton... On complete securities markets and the martingale property on securities prices. Applied Probability 2. E. The valuation of American puts with exponential exercise policies. The pricing of the American option. S. 145–173. J.362 6. M. Continuous-Time Finance. V. New York. J. 100–128. Business 50. Cambridge. C. Springer-Verlag. . F. Ann. (1996) Musiela. P. Econ. 1–23. (1973) Merton. R. Management Science 4. M. (1977) Parkinson. Review Economics Statistics 51. Reprinted in Cootner (1964). Adv.. (1995) Øksendal. C. Futures Option Research 2. Bell J. R. (1976) Merton. Stochastic Differential Equations. M. 1990... (1989) M¨ ller. Theory and Applications. C. Brownian motion in the stock market. 34–35. Springer-Verlag. Theory of rational option pricing. (1992) Myneni. 4th Edition. If the cap fits . R.. New York.

Inst. Math. Probab. . Strategy and Applications. & Sundaresan. W. Proof that properly anticipated prices fluctuate randomly. (1995) Rogers. (1992b) Schweizer. henri Poincar´ 26. J. (1990) Shirakawa. 77–94. G. Yale Economic Essays 1. New York. 65: Mathematical Finance. M. Martingale densities for general asset prices. Adv. D. (1987) Ritchken. Financial Economics 5. P. The analysis of finite security markets using martingales. M. Math. 2. (1991) Shirakawa. Warrant prices as indicators of expectations and preferences. F.. J. & Willinger. Tokyo Institute of Technology. 451–460. (1973) Samuelson. (1965) Samuelson. (1992a) Schweizer. M. Reprinted in Cootner (1964). (M. preprint IHSS 90-20. (1990) Stricker.. IMA Vol. P. e (1987) Taqqu.. 9. L. Probab. 28–35. Which model of term-structure of interest rates should one use?. Ann. Optimal dividend and portfolio decisions with Poisson and diffusion-type return processes. Appl. An equilibrium characterization of the term structure. P. 178–231. Duffie.. 177–188. 1–42. Fleming & S. Economics 21.. 41–50.. 1–25.. The value of an Asian option. Mathematics of speculative prices. editors). M. Risk Magazine 4/8.CHAPTER 35. 93–116.. M. 19. (1991) Rubinstein. Notes and References 363 (1981) Richard. M. Breaking down the barriers.. E. & Shi. C. Options: Theory. Finance 1(4). 123– 131. L. 347–371. G. Davis. Shreve. W. Arbitrage et lois de martingales. C. SIAM Review 15. 412–474. to appear. A continuous-time equilibrium model of forward and futures prices in a multigood economy. Applied Probability. J. S.. C. & Reiner. H. C. Financial Economics. (1977) Vasicek. J. Harper Collins. H... A. 171– 179. O. Mean-variance hedging for general claims.. (1961) Sprenkle. Z.. Annals Appl. A. Interest rate option pricing with Poisson-Gaussian forward rate curve processes. Industrial Management Review 4. Springer-Verlag. (1995) Rogers.

Dewynne.. . P. Oxford Financial Press. Cambridge University Press. Bull. K. France 52. Math. Cambridge. Insurance: Mathematics and Economics 14. (1992) Xue.. (1991) Williams. Differential spaces. Probability with Martingales. X. The Mathematics of Financial Derivatives. (1993) Wilmott. J... Physics 2. & Howison. 279–311. S. Math. J. D. Cambridge.. ´ (1939) Ville. N. PO Box 348. 139–161. in Lecture Notes in Control and Information Sciences 177.364 (1994) Vetzal. Oxford. Option Pricing. 131–174. OX4 1DR. Dewynne. Etude Critique de la Notion du Collectif.. Cambridge University Press. A survey of stochastic continuous time models of the term structure of interest rates.. (1924) Wiener. P. Gauthier-Villars. Berlin.. Soc. (1923) Wiener. 569– e ´ e 578. and its applications. R. (1995) Wilmott. Paris. J. SpringerVerlag. & Howison. Un probl´ me de probabilit es d´ nombrables. UK. J. N. S. Martingale representation for a class of processes with independent increments... X.

Sign up to vote on this title
UsefulNot useful

Master Your Semester with Scribd & The New York Times

Special offer: Get 4 months of Scribd and The New York Times for just $1.87 per week!

Master Your Semester with a Special Offer from Scribd & The New York Times