American-type options. Stochastic approximation methods. Volume 1 (Q384891)

In this book, the author presents various stochastic approximation methods employed in modeling American-type options with general pay-off functions for discrete time Markov log-price processes. The pay-off functions may depend not only on the price, but also upon an additional stochastic index component with a general phase space. The book is structured in ten chapters. Chapter 1 introduces models of multivariate modulated Markov log-price processes and price processes. Various types of modulations are considered here. Some examples of such processes are presented, such as: the log-price processes represented by multivariate modulated random walks, the log-price processes represented by various autoregressive and autoregressive moving average type and the autoregressive and autoregressive moving average stochastic volatility models. The second chapter presents the American-type options and introduces the basic objects connected with the American-type options for multivariate modulated Markov price and log-price processes. Among these objects we mention: the pay-off functions, reward and log-reward functions, optimal expected rewards, optimal stopping times etc. Basic optimization problems for American-type options are formulated. In the third chapter the author investigates the structure of backward recurrence relations for log-price processes represented by atomic Markov chains having the transition probabilities concentrated on finite sets. The chapter discusses in detail the univariate and multivariate binomial tree reward algorithms, trinomial tree reward algorithms, the random walk reward algorithms and the Markov chain reward algorithms. Chapter 4 deals with upper bounds for log-reward functions and optimal expected rewards for American-type functions. The models considered here have pay-off functions having not more than polynomial rate of growth in price arguments. The next two chapters contain convergence results for rewards of American-type options with general perturbed pay-off functions for perturbed multivariate modulated Markov log-price processes or with bounded and unbounded characteristics. Minimal conditions of smoothness on the limiting pay-off functions and transition probabilities are imposed. Chapter 7 contains general results about the so-called space-skeleton approximations for rewards in the case of general pay-off functions. The new results in this chapter are related to the convergence of space-skeleton approximations for rewards of American-type options for log-price processes represented by multivariate modulated random walks and of American-type options for for multivariate modulated Markov log-price processes with unbounded characteristics. In Chapter 8, the author presents results about the convergence of option rewards and space-skeleton approximations for multivariate modulated Markov Gaussian log-price processes with either bounded or unbounded drift and volatility coefficients. The last two chapters of the book contain results about fitting the coefficients and convergence results for reward functions for univariate and multivariate binomial and trinomial approximations. The cases of homogeneity and inhomogeneity in space and time are considered. The book ends with some bibliographical remarks and a comprehensive list of references.