Dans cette thèse, nous traitons deux problèmes de contrôle optimal stochastique. Chaque problème correspond à une Partie de ce document. Le premier problème traité est très précis, il s'agit de la valorisation des contrats optionnels de vente de type Américain (dit Put Américain) en présence de dividendes discrets (Partie I). Le deuxième est plus général, puisqu'il s'agit dans un cadre discret en temps de prouver l'existence d'un principe de programmation dynamique sous des contraintes en probabilités (Partie II). Bien que les deux problèmes soient assez distincts, le principe de programmation dynamique est au coeur de ces deux problèmes. La relation entre la valorisation d'un Put Américain et un problème de frontière libre a été prouvée par McKean. La frontière de ce problème a une signification économique claire puisqu'elle correspond à tout instant à la borne supérieure de l'ensemble des prix d'actifs pour lesquels il est préférable d'exercer tout de suite son droit de vente. La forme de cette frontière en présence de dividendes discrets n'avait pas été résolue à notre connaissance. Sous l'hypothèse que le dividende est une fonction déterministe du prix de l'actif à l'instant précédant son versement, nous étudions donc comment la frontière est modifiée. Au voisinage des dates de dividende, et dans le modèle du Chapitre 3, nous savons qualifier la monotonie de la frontière, et dans certains cas quantifier son comportement local. Dans le Chapitre 3, nous montrons que la propriété du smooth-fit est satisfaite à toute date sauf celles de versement des dividendes. Dans les deux Chapitres 3 et 4, nous donnons des conditions pour garantir la continuité de cette frontière en dehors des dates de dividende. La Partie II est originellement motivée par la gestion optimale de la production d'une centrale hydro-electrique avec une contrainte en probabilité sur le niveau d'eau du barrage à certaines dates. En utilisant les travaux de Balder sur la relaxation de Young des problèmes de commande optimale, nous nous intéressons plus spécifiquement à leur résolution par programmation dynamique. Dans le Chapitre 5, nous étendons au cadre des mesures de Young des résultats dûs à Evstigneev. Nous établissons alors qu'il est possible de résoudre par programmation dynamique certains problèmes avec des contraintes en espérances conditionnelles. Grâce aux travaux de Bouchard, Elie, Soner et Touzi sur les problèmes de cible stochastique avec perte contrôlée, nous montrons dans le Chapitre 6 qu'un problème avec contrainte en espérance peut se ramener à un problème avec des contraintes en espérances conditionnelles. Comme cas particulier, nous prouvons ainsi que le problème initial de la gestion du barrage peut se résoudre par programmation dynamique / In this thesis, we address two problems of stochastic optimal control. Each problem constitutes a different Part in this document. The first problem addressed is very precise, it is the valuation of American contingent claims and more specifically the American Put in the presence of discrete dividends (Part I). The second one is more general, since it is the proof of the existence of a dynamic programming principle under expectation constraints in a discrete time framework (Part II). Although the two problems are quite distinct, the dynamic programming principle is at the heart of these two problems. The relationship between the value of an American Put and a free boundary problem has been proved by McKean. The boundary of this problem has a clear economic meaning since it corresponds at all times to the upper limit of the asset price above which the holder of such an option would exercise immediately his right to sell. The shape of the boundary in the presence of discrete dividends has not been solved to the best of our knowledge. Under the assumption that the dividend is a deterministic function of asset prices at the date just before the dividend payment, we investigate how the boundary is modified. In the neighborhood of dividend dates and in the model of Chapter 3, we know what the monotonicity of the border is, and we quantify its local behavior. In Chapter 3, we show that the smooth-fit property is satisfied at any date except for those of the payment of dividends. In both Chapters 3 and 4, we are able to give conditions to guarantee the continuity of the border outside dates of dividend. Part II was originally motivated by the optimal management of the production of an hydro-electric power plant with a probability constraint on the reservoir level on certain dates. Using Balder'sworks on Young's relaxation of optimal control problems, we focus more specifically on their resolution by dynamic programming. In Chapter 5, we extend results of Evstigneev to the framework of Young measures. We show that dynamic programming can be used to solve some problems with conditional expectations constraints. Through the ideas of Bouchard, Elie, Soner and Touzi on stochastic target problems with controlled loss, we show in Chapter 6 that a problem with expectation constraints can be reduced to a problem with conditional expectation constraints. Finally, as a special case, we show that the initial problem of dam management can be solved by dynamic programming

Les marchés financiers ont connu, grâce aux études réalisées durant les trois dernières décennies, une expansion considérable et ont vu l’apparition de produits dérivés divers et variés. Les plus utilisés parmi ces produits dérivés sont les options américaines / Financial markets knew, thanks to studies carried out during the last three decades, a considerable expansion and saw the appearance of diverse and varied by-products. The most used among these by-products are the American options

In der Finanzmathematik hat der Besitzer einer amerikanische Option das Recht aber nicht die Pflicht, eine Aktie innerhalb eines bestimmten Zeitraums, für einen bestimmten Preis zu kaufen oder zu verkaufen. Die Bewertung einer amerikanische Option wird als so genanntes optimale stopping Problem formuliert. Erfolgt die Modellierung des Aktienkurses durch eine geometrische Brownsche Bewegung, wird der Wert einer amerikanischen Option durch ein deterministisches freies Randwertproblem (FRWP), oder einer äquivalenten Variationsungleichung (VU) auf ganz R in gewichteten Sobolev Räumen gegeben. Um Standardmethoden der Numerischen Mathematik anzuwenden, wird das unbeschränkte Gebiet zu einem beschränkten Gebiet abgeschnitten. Mit Hilfe der Fourier-Transformation wird eine Integraldarstellung der Lösung die den freien Rand explizit beinhaltet, hergeleitet. Mittels dieser Integraldarstellung werden Abschneidefehlerschranken bewiesen. Danach werden gewichtete Poincare Ungleichungen mit expliziten Konstanten bewiesen. Der Abschneidefehler und die gewichtete Poincare Ungleichung ermöglichen, einen zuverlässigen a posteriori Fehlerschätzer zwischen der exakten Lösung der VU und der semidiskreten Lösung des penalisierten Problems auf R herzuleiten. Eine hinreichend glatte Lösung der VU garantiert die Konvergenz der Lösung des penaltisierten Problems zur Lösung der VU. Ein a priori Fehlerschätzer für den Fehler zwischen der exakten Lösung der VU und der semidiskreten Lösung des penaltisierten Problems beendet die numerische Analysis. Die eingeführten aposteriori Fehlerschätzer motivieren einen Algorithmus für adaptive Netzverfeinerung. Numerische Experimente zeigen die verbesserte Konvergenz des adaptiven Verfahrens gegenüber der uniformen Verfeinerung. Der zuverlässige a posteriori Fehlerschätzer ermöglicht es, den Abschneidepunkt so zu wählen, dass der Gesamtfehler (Diskretisierungsfehler plus Abschneidefehler) kleiner als eine gegebenen Toleranz ist. / Among the central concerns in mathematical finance is the evaluation of American options. An American option gives the holder the right but not the obligation to buy or sell a certain financial asset within a certain time-frame, for a certain strike price. The valuation of American options is formulated as an optimal stopping problem. If the stock price is modelled by a geometric Brownian motion, the value of an American option is given by a deterministic parabolic free boundary value problem (FBVP) or equivalently a non-symmetric variational inequality (VI) on weighted Sobolev spaces on R. To apply standard numerical methods, the unbounded domain R is truncated to a bounded one. Applying the Fourier transform to the FBVP yields an integral representation of the solution including the free boundary explicitely. This integral representation allows to prove explicit truncation errors. Since the VI is formulated within the framework of weighted Sobolev spaces, we establish a weighted Poincare inequality with explicit determined constants. The truncation error estimate and the weighted Poncare inequality enable a reliable a posteriori error estimate between the exact solution of the VI and the semi-discrete solution of the penalised problem on R. A sufficient regular solution provides the convergence of the solution of the penalised problem to the solution of the VI. An a priori error estimate for the error between the exact solution of the VI and the semi-discrete solution of the penalised problem concludes the numerical analysis. The established a posteriori error estimates motivates an algorithm for adaptive mesh refinement. Numerical experiments show the improved convergence of the adaptive algorithm compared to uniform mesh refinement. The reliable a posteriori error estimate including explicit truncation errors allows to determine a truncation point such that the total error (discretisation and truncation error) is below a given error tolerance.

L’objet de cette thèse est l’étude de problèmes d’évaluation d’options dans les modèles à volatilité stochastique. La première partie est centrée sur les options américaines dans le modèle de Heston. Nous donnons d’abord une caractérisation analytique de la fonction de valeur d’une option américaine comme l’unique solution du problème d’obstacle parabolique dégénéré associé. Notre approche est basée sur des inéquations variationelles dans des espaces de Sobolev avec poids étendant les résultats récents de Daskalopoulos et Feehan (2011, 2016) et Feehan et Pop (2015). On étudie aussi les propriétés de la fonction de valeur d’une option américaine. En particulier, nous prouvons que, sous des hypothèses convenables sur le payoff, la fonction de valeur est décroissante par rapport à la volatilité. Ensuite nous nous concentrons sur le put américaine et nous étendons quelques résultats qui sont bien connus dans le monde Black-Scholes. En particulier nous prouvons la convexité stricte de la fonction de valeur dans la région de continuation, quelques propriétés de la frontière libre, la formule de Prime d’Exercice Anticipée et une forme faible de la propriété du smooth fit. Les techniques utilisées sont de type probabiliste. Dans la deuxième partie nous abordons le problème du calcul numérique du prix des options européennes et américaines dans des modèles à volatilité stochastiques et avec sauts. Nous étudions d’abord le modèle de Bates-Hull-White, c’est-à-dire le modèle de Bates avec un taux d’intérêt stochastique. On considère un algorithme hybride rétrograde qui utilise une approximation par chaîne de Markov (notamment un arbre “avec sauts multiples”) dans la direction de la volatilité et du taux d’intérêt et une approche (déterministe) par différence finie pour traiter le processus de prix d’actif. De plus, nous fournissons une procédure de simulation pour des évaluations Monte Carlo. Les résultats numériques montrent la fiabilité et l’efficacité de ces méthodes. Finalement, nous analysons le taux de convergence de l’algorithme hybride appliqué à des modèles généraux de diffusion avec sauts. Nous étudions d’abord la convergence faible au premier ordre de chaînes de Markov vers la diffusion sous des hypothèses assez générales. Ensuite nous prouvons la convergence de l’algorithme: nous étudions la stabilité et la consistance de la méthode hybride par une technique qui exploite les caractéristiques probabilistes de l’approximation par chaîne de Markov / We study option pricing problems in stochastic volatility models. In the first part of this thesis we focus on American options in the Heston model. We first give an analytical characterization of the value function of an American option as the unique solution of the associated (degenerate) parabolic obstacle problem. Our approach is based on variational inequalities in suitable weighted Sobolev spaces and extends recent results of Daskalopoulos and Feehan (2011, 2016) and Feehan and Pop (2015). We also investigate the properties of the American value function. In particular, we prove that, under suitable assumptions on the payoff, the value function is nondecreasing with respect to the volatility variable. Then, we focus on an American put option and we extend some results which are well known in the Black and Scholes world. In particular, we prove the strict convexity of the value function in the continuation region, some properties of the free boundary function, the Early Exercise Price formula and a weak form of the smooth fit principle. This is done mostly by using probabilistic techniques.In the second part we deal with the numerical computation of European and American option prices in jump-diffusion stochastic volatility models. We first focus on the Bates-Hull-White model, i.e. the Bates model with a stochastic interest rate. We consider a backward hybrid algorithm which uses a Markov chain approximation (in particular, a “multiple jumps” tree) in the direction of the volatility and the interest rate and a (deterministic) finite-difference approach in order to handle the underlying asset price process. Moreover, we provide a simulation scheme to be used for Monte Carlo evaluations. Numerical results show the reliability and the efficiency of the proposed methods.Finally, we analyze the rate of convergence of the hybrid algorithm applied to general jump-diffusion models. We study first order weak convergence of Markov chains to diffusions under quite general assumptions. Then, we prove the convergence of the algorithm, by studying the stability and the consistency of the hybrid scheme, in a sense that allows us to exploit the probabilistic features of the Markov chain approximation

In dieser Dissertation werden Amerikanischen Optionen in einem unvollst¨andigen Markt und in stetiger Zeit untersucht. Die Dissertation besteht aus zwei Teilen. Im ersten Teil untersuchen wir ein stochastisches Optimierungsproblem, in dem ein konvexes robustes Verlustfunktional ueber einer Menge von stochastichen Integralen minimiert wird. Dies Problem tritt auf, wenn der Verkaeufer einer Amerikanischen Option sein Ausfallsrisiko kontrollieren will, indem er eine Strategie der partiellen Absicherung benutzt. Hier quantifizieren wir das Ausfallsrisiko durch ein robustes Verlustfunktional, welches durch die Erweiterung der klassischen Theorie des erwarteten Nutzens durch Gilboa und Schmeidler motiviert ist. In einem allgemeinen Semimartingal-Modell beweisen wir die Existenz einer optimalen Strategie. Unter zusaetzlichen Kompaktheitsannahmen zeigen wir, wie das robuste Problem auf ein nicht-robustes Optimierungsproblem bezueglich einer unguenstigsten Wahrscheinlichkeitsverteilung reduziert werden kann. Im zweiten Teil untersuchen wir die obere und die untere Snellsche Einhuellende zu einer Amerikanischen Option. Wir konstruieren diese Einhuellenden fuer eine stabile Familie von aequivalenten Wahrscheinlichkeitsmassen; die Familie der aequivalentenMartingalmassen ist dabei der zentrale Spezialfall. Wir formulieren dann zwei Probleme des robusten optimalen Stoppens. Das Stopp-Problem fuer die obere Snellsche Einhuellende ist durch die Kontrolle des Risikos motiviert, welches sich aus der Wahl einer Ausuebungszeit durch den Kaeufer bezieht, wobei das Risiko durch ein kohaerentes Risikomass bemessen wird. Das Stopp-Problem fuer die untere Snellsche Einhuellende wird durch eine auf Gilboa und Schmeidler zurueckgehende robuste Erweiterung der klassischen Nutzentheorie motiviert. Mithilfe von Martingalmethoden zeigen wir, wie sich optimale Loesungen in stetiger Zeit und fuer einen endlichen Horizont konstruieren lassen. / This thesis studies American options in an incomplete financial market and in continuous time. It is composed of two parts. In the first part we study a stochastic optimization problem in which a robust convex loss functional is minimized in a space of stochastic integrals. This problem arises when the seller of an American option aims to control the shortfall risk by using a partial hedge. We quantify the shortfall risk through a robust loss functional motivated by an extension of classical expected utility theory due to Gilboa and Schmeidler. In a general semimartingale model we prove the existence of an optimal strategy. Under additional compactness assumptions we show how the robust problem can be reduced to a non-robust optimization problem with respect to a worst-case probability measure. In the second part, we study the notions of the upper and the lower Snell envelope associated to an American option. We construct the envelopes for stable families of equivalent probability measures, the family of local martingale measures being an important special case. We then formulate two robust optimal stopping problems. The stopping problem related to the upper Snell envelope is motivated by the problem of monitoring the risk associated to the buyer’s choice of an exercise time, where the risk is specified by a coherent risk measure. The stopping problem related to the lower Snell envelope is motivated by a robust extension of classical expected utility theory due to Gilboa and Schmeidler. Using martingale methods we show how to construct optimal solutions in continuous time and for a finite horizon.

O objetivo desta dissertação é fornecer primeiramente o arcabouço necessário para o entendimento do derivativo opções, muito utilizado nos mercados financeiros mundiais, e posteriormente executar precificações de opções americanas a partir dos modelos dos mínimos quadrados de Monte Carlo (LSM), o modelo de árvore binomial com extrapolação de Richardson e a aproximação analítica de Bjerksund e Stensland (B&S), aplicando duas plataformas de processamento paralelo computacional, a TPL (Task Parallel Library) nativa no .NET framework 4.5 e a plataforma CUDA (Compute Unified Device Architecture), demonstrando o comparativo dos resultados obtidos a cada modelo diante de cada plataforma. / The objective of this dissertation is to provide first the necessary framework for the understanding of the derivative options, widely used in the world financial markets, and later to execute the American option pricing from Monte Carlo least squares models (LSM), the binomial tree model with Richardson extrapolation and the Bjerksund and Stensland analytic approach (BJS) by applying two parallel computational processing platforms, the native TPL (Task Parallel Library) in the .NET framework 4.5 and the CUDA platform (Compute Unified Device Architecture), demonstrating the comparison of the obtained results to each model before each platform.

<p>This thesis consists of four papers and a summary. The common topic of the included papers are the pricing equations of path-dependent options. Various properties of barrier options and American options are studied, such as convexity of option prices, the size of the continuation region in American option pricing and pricing formulas for turbo warrants. In Paper I we study the effect of model misspecification on barrier option pricing. It turns out that, as in the case of ordinary European and American options, this is closely related to convexity properties of the option prices. We show that barrier option prices are convex under certain conditions on the contract function and on the relation between the risk-free rate of return and the dividend rate. In Paper II a new condition is given to ensure that the early exercise feature in American option pricing has a positive value. We give necessary and sufficient conditions for the American option price to coincide with the corresponding European option price in at least one diffusion model. In Paper III we study parabolic obstacle problems related to American option pricing and in particular the size of the non-coincidence set. The main result is that if the boundary of the set of points where the obstacle is a strict subsolution to the differential equation is C¹-Dini in space and Lipschitz in time, there is a positive distance, which is uniform in space, between the boundary of this set and the boundary of the non-coincidence set. In Paper IV we derive explicit pricing formulas for turbo warrants under the classical Black-Scholes assumptions.</p>

This thesis consists of four papers and a summary. The common topic of the included papers are the pricing equations of path-dependent options. Various properties of barrier options and American options are studied, such as convexity of option prices, the size of the continuation region in American option pricing and pricing formulas for turbo warrants. In Paper I we study the effect of model misspecification on barrier option pricing. It turns out that, as in the case of ordinary European and American options, this is closely related to convexity properties of the option prices. We show that barrier option prices are convex under certain conditions on the contract function and on the relation between the risk-free rate of return and the dividend rate. In Paper II a new condition is given to ensure that the early exercise feature in American option pricing has a positive value. We give necessary and sufficient conditions for the American option price to coincide with the corresponding European option price in at least one diffusion model. In Paper III we study parabolic obstacle problems related to American option pricing and in particular the size of the non-coincidence set. The main result is that if the boundary of the set of points where the obstacle is a strict subsolution to the differential equation is C1-Dini in space and Lipschitz in time, there is a positive distance, which is uniform in space, between the boundary of this set and the boundary of the non-coincidence set. In Paper IV we derive explicit pricing formulas for turbo warrants under the classical Black-Scholes assumptions.

[pt] A maioria das opções negociadas atualmente é do estilo americano, no entanto sua avaliação continua sendo uma tarefa bastante difí­cil, constituindo-se numa das áreas mais desafiadoras no campo de derivativos financeiros, particularmente quando existem vários fatores afetando o preço da opção. Isso ocorre basicamente porque os métodos de árvores binomiais e diferenças finitas tornam-se impraticáveis na avaliação de opções com mais de três fatores de incerteza. No presente trabalho, faz-se um estudo prévio dos modelos de precificação tradicionais, para posteriormente nos estendermos a modelos mais flexíveis desenvolvidos recentemente baseados em simulações de Monte Carlo e Quase-Monte Carlo, até então considerados inaplicáveis na avaliação de opções americanas. Nesse sentido, pretendemos comprovar a aplicabilidade e versatilidade dos modelos baseados em simulação na avaliação de opções americanas tradicionais ou complexas. Nossa análise baseia-se, sobretudo na ilustração de exemplos práticos, dando especial ênfase à implementação computacional e precisão dos modelos. / [en] The majority of the options negotiated nowadays are of the american style, however its valuation goes on being a very hard job, constituting themselves in one of the most challenging areas in the financial derivative field, particularly when there are several factors affecting the price of the option. It happens basically because the binominal trees and finite differences methods become impracticable in the valuation of options with more than three factors of uncertainty. In this work we are doing a previous study of the traditional methods of american option valuation for later extending this study to more flexible and newly developed models based on simulations of Monte Carlo and Quase-Monte Carlo, which up to the present have been considered inapplicable in the valuation of the american style options. In this sense we intend to prove the applicability and versatility of the models based on simulation in the valuation of traditional and complex american options. Our analysis is, above all based on the illustration of practical examples giving special emphasis to the computational implementation and accuracy of the methods.

O objetivo desta dissertação é fornecer primeiramente o arcabouço necessário para o entendimento do derivativo opções, muito utilizado nos mercados financeiros mundiais, e posteriormente executar precificações de opções americanas a partir dos modelos dos mínimos quadrados de Monte Carlo (LSM), o modelo de árvore binomial com extrapolação de Richardson e a aproximação analítica de Bjerksund e Stensland (B&S), aplicando duas plataformas de processamento paralelo computacional, a TPL (Task Parallel Library) nativa no .NET framework 4.5 e a plataforma CUDA (Compute Unified Device Architecture), demonstrando o comparativo dos resultados obtidos a cada modelo diante de cada plataforma. / The objective of this dissertation is to provide first the necessary framework for the understanding of the derivative options, widely used in the world financial markets, and later to execute the American option pricing from Monte Carlo least squares models (LSM), the binomial tree model with Richardson extrapolation and the Bjerksund and Stensland analytic approach (BJS) by applying two parallel computational processing platforms, the native TPL (Task Parallel Library) in the .NET framework 4.5 and the CUDA platform (Compute Unified Device Architecture), demonstrating the comparison of the obtained results to each model before each platform.