Mathematical Finance, Spring 2021
Dario Gasbarra1
March 26, 2021
1University of Helsinki|Master program in Mathematics and Statis- tics|MAST31801 2
My Thesis, paradoxically and a little provocatively, but nonetheless gen- uinely, is simply this:
PROBABILITY DOES’ NOT EXISTS.
The abandonment of superstitious beliefs about the existence of Phlogis- ton , the Cosmic Ether, Absolute Space and Time, ... , or Faires and Witches was an essential step along the road on scientific thinking. Probability too, if regarded as something endowed with some kind of objective existence is no less a misleading misconception, an illusory attempt to exteriorize or mate- rialize our true probabilitic beliefs.
Bruno De Finetti, Theory of Probability (1974). Contents
1 Probability = Price 5 1.0.1 Expectation ...... 9 1.1 Convexity and separating hyperplane theorem ...... 11 1.1.1 First fundamental Theorem from Farkas lemma . . . . 15 1.1.2 Separation theorem in infinite dimension ...... 20 1.2 Kolmogorov axioms of probability ...... 20 1.2.1 Lebesgue integral and expectation ...... 22 1.3 Arbitrage in one period market model...... 25 1.3.1 Discounting ...... 26 1.4 Geometric Characterization of Arbitrage Free Markets . . . . 30 1.5 First Fundamental Theorem in Multiperiod models ...... 33 1.6 Change of numeraire ...... 40 1.7 Option pricing ...... 41 1.7.1 Intermezzo: The Gaussian integration by parts for- mula ...... 43 1.7.2 Black and Scholes european call option pricing formula for the lognormal stock price ...... 44 1.7.3 Hedging and superhedging contingent claims . . . . . 47 1.8 Convex functions ...... 54 1.8.1 Replicating european options using call options . . . . 56 1.8.2 Leverage effect of option investments ...... 57
2 Martingale representation and hedging 61 2.1 Some basic facts from martingale theory ...... 61 2.1.1 Conditional Expectation and Martingales ...... 61 2.1.2 Square integrable martingales and predictable bracket . 66 2.1.3 Orthogonal projections in the space of square inte- grable martingales ...... 67 2.1.4 Martingale property and change of measure ...... 68 2.1.5 Doob decomposition and change of measure ...... 69 2.1.6 Stopping times and localization ...... 71
3 4 CONTENTS
2.1.7 Martingale Predictable Representation Property . . . 73 2.2 Uniform Doob decomposition ...... 76
3 Option Pricing and Hedging in the Cox-Ross-Rubinstein Bi- nomial market model 81
4 American options in discrete time 91 4.1 Optimal Stopping and dynamic programming ...... 91 4.2 Pricing and hedging american options ...... 93 4.2.1 Hedging american options ...... 94 4.2.2 Optimal exercise strategies for the option holder . . . 95 4.2.3 Relation between american and european call options . 95 4.3 Complements on American option ...... 96 4.4 Dual representation ...... 97 4.5 Software tools for computing some option prices in the bino- mial tree model ...... 98
5 Calculus with respect to functions of finite variation 103
6 Convergence towards time continuous Black & Scholes model 111
7 Quadratic variation and Ito-Föllmer calculus 115 7.0.1 Ito-Föllmer calculus for random paths ...... 124 7.0.2 Cross-variation ...... 128 7.0.3 Pathwise Stratonovich calculus ...... 130 Chapter 1
Probability = Price
Bruno De Finetti (1906-1985) was an italian mathematician, economist and philosopher. In his philosophy of science the postulate of absolute proba- bility is abandoned. Instead probabilities have a purely operative meaning. Probabilities are always relative to our state of knowledge/ignorance.
In an uncertain world, we classify events into two types C = { certain events } and U = { uncertain events }. When E ∈ C is certain and F ⊇ E, meaning that also F happens when E happens, it follows that also F ∈ C is a certain event. We say that an event E is impossible when its complement or negation Ec (which happens when E does not happen) is a certain event.
We may say that uncertain events which are not impossible are random events. If E and F are events, we denote by (E ∪ F ) the event where either E or F or perhaps both of them occur, and (E ∩ F ) the event where both E and F occur,
We denote the class N = { impossible events } and we say that two events E,F are not compatible or disjoint if their joint occurrence is an impossible event. eli (E ∩F ) ∈ N . Note also that for any event E, always (E ∪Ec) ∈ C.
A bookmaker accepts bets from his customers on the occurence of the disjoint events E1,...,En, such that (Ei ∩ Ej) ∈ N (impossible event) for i 6= j.
The bookmaker chooses the prices of the gambles for the events Ei, their
5 6 CHAPTER 1. PROBABILITY = PRICE
c complements Ei , and all possible unions (Ei ∪ Ej), (Ei ∪ Ej ∪ Ek),... as
Pr(Ei), Pr(Ei ∪ Ej), Pr(Ei ∪ Ej ∪ Ek) · · · ∈ R.
Remark 1.0.1. The set of prices for all the considered events is called pricing system. The notation Pr is both for Prices and Probabilities. Gamblers are allowed to choose their gambling strategy and place simul- taneously gambles against all these considered events A with both positive and negative coefficients (for each event both long and short position are allowed), and the bookmaker has to accept every possible bet at the prices he has set. Let’s introduce the indicator of an event 1A which takes value 1 if A occurs and 0 if A does not occur. Such indicators are random variables, whose values is not specified in advance but it depends on the occurrence of the random event A. A customer is gambling against the events E1,...En with a gambling n strategy y = (y1, . . . , yn) ∈ R of his choice. This means that he pays
y1Pr(E1) + y2Pr(E2) + ··· + ynPr(En) and received the value of the random variable
y11E1 + y21E2 + ··· + yn1En with profit
V = (1E1 − p1)y1 + ··· + (1E1 − pn)yn where pi = Pr(Ei). A negative profit corresponds to a loss.
Definition 1.0.1. An arbitrage opportunity is a situation where by following a suitable strategy the customer makes a profit which is certainly non-negative and it is possible (which means that it is not impossible) that the profit is strictly positive. If the bookmaker wants to stay in business, he must choose the gambling prices avoiding to create arbitrage opportunities for his customers. We show that in order to do so, the bookmaker has to be coherent and use a probability to form the price system.
Theorem 1.0.1. A pricing system free of arbitrage must be a finitely additive probability, with the following properties: 7
1. The price of the gambles are unique (law of one price), and every the prices of event indicators 1E are numbers in the interval [0, 1]. 2. If E ∈ C is certain, necessarily P r(E) = 1 and if E ∈ N is impossible, necessarily P r(E) = 0. 3. Prices of events are (finitely) additive.
Proof. 1. If the bet 1E had two prices, p1 > p2, a gambler would place an arbitrary large number of bets x > 0 at price p1 and simulataneously place (−x) bets at price p2. Trivially this is an arbitrage strategy, where gambler’s profit (which is the loss of the bookmaker) in all situations with certainity is given by
(1E − p1)x − (1E − p2)x = (p1 − p2)x
2. If the event E is certain, and the bet 1E has price p 6= 1, the gambler placing x bets makes a profit
(1E − p)x = (1 − p)x with certainity. When p > 1 ( p < 1) the gambler will make with cer- tainity an arbitrary large profit with x < 0 ( x > 0). If E is impossible, with certainity the gambler profit is
(1E − p)x = −px where again the gambler chooses an arbitrary large x < 0. 3. We consider first the situation where we price two disjoint events, n = 2 , (E1 ∩ E2) ∈ N , such that their union (E1 ∪ E2) ∈ C is a certain event. If the pricing system is arbitrage free, we should have P r(E1 ∪E2) = 1. The linear system V (E1) = (1 − p1)y1 − p2y2 = gambler’s profit when E1 occurs V (E2) = −p1y1 + (1 − p2)y2 = gambler’s profit when E2 occurs
has a solution (y1, y2) for any profit vector (V (E1),V (E2)) if and only if the system coefficients matrix is invertible, 1 − p1 −p2 det = 1 − p1 − p2 6= 0 −p1 1 − p2 In order to avoid arbitrage possibilities for the gamblers, this determi- nant has to be zero, equivalently
P r(E1) + P r(E2) = 1 = P r(E1 ∪ E2) = 1. 8 CHAPTER 1. PROBABILITY = PRICE
Similarly, for n > 2 disjoint events Ei with Ei ∩ Ej ∈ N for i 6= j, such that their union E1 ∪ E2 ∪ · · · ∪ En ∈ C is a certain event , the price system pi = Pr(Ei), i = 1, . . . , n is arbitrage-free if and only if 1 − p1 −p2 ... −pn −p1 1 − p2 ... −pn det = 1 − p1 − p2 − · · · − pn = 0 ...... −p1 −p2 ... 1 − pn
This determinant is computed by induction or by using Sylvester lemma from linear algebra:
Lemma 1.0.1. If A and B are respectively n × m and m × n matrices, and In denotes the n × n identity matrix, det In + AB = det Im + BA
More in general, for two disjoint events (E1 ∩ E2) ∈ N ,