Peacocks and associated martingales, with explicit constructions (Q538823)

A typical problem in mathematical finance is how to determine option prices. That is, assuming that a stochastic process \(X\) modelling some asset price is given, we try to find the ``correct'' price, e.g., for the European call option \((X_T - K)^+\). It turns out that a price \(p\) does not introduce any arbitrage into the market if and only if \(p ={\text E}_Q(X_T - K)^+\), where \(Q\) is an equivalent martingale measure. In this monograph, the authors take a somewhat opposite point of view. They assume that only the one-dimensional marginals law\((X_t)\) for \(t \geq 0\) are given. In particular, the prices of European call and put options are determined. They then try to construct a continuous-time model which reproduces these option prices. By arbitrage considerations, this must be a martingale \((M_t: t \geq 0)\) such that \[ \text{law}(M_t) =\text{law}(X_t) \] for every \(t \geq 0\). It turns out that this is possible if and only if the family of one-dimensional marginals \((X_t: t \geq 0)\) is increasing in the convex order (Processus Croissant pour l'Ordre Convexe, which becomes PCOC and then peacock -- hence the title). The sufficiency of the peacock property is a deep result of Kellerer, which sadly is not proven in the monograph -- especially since the original article is written in German; this would have been a good occasion to make it available to a broader audience. However, the proof of Kellerer is nonconstructive. Therefore, the program of the book can rougly be described as follows: First, find sufficiently many peacocks. Then, try to explicitly construct a martingale associated to a given peacock. The first and longest chapter deals with the first part of the program: The authors provide several characterizations of the peacock property, and then continue to provide many different examples of peacocks. The ``guiding example'' of a peacock is the running average of the geometric Brownian motion: \[ \left(\tfrac{1}{t} \int_0^t e^{B_s - \frac{s}{2}}\,ds,\quad t\geq 0\right). \] This is generalized in many different ways. But the authors also provide a large number of peacocks not derived from the guiding example. It should be pointed out that, in the first chapter alone, there are more than 40 exercises. The following six chapters are devoted to the second part of the program: Construct a martingale associated to a given peacock. Each chapter presents a different method. The presented methods are: Chapter 2: The sheet method; Chapter 3: The time reversal method; Chapter 4: The time inversion method; Chapter 5: The Sato process method; Chapter 6: The stochastic differential equation method; Chapter 7: The Skorokhod embedding (SE) method. The eighth and last chapter asks how to compare different martingales associated to a given peacock \(X\). In particular, the authors compare the \(k\)-dimensional margins law\((M_{t_1}, \dots, M_{t_k})\) for \(0 \leq t_1 < \dots < t_k\) in the upper orthant order. The monograph ends with a list of open problems. Throughout the text there are lots of exercises (about 100) and examples. The main target groups are clearly researchers and practitioners. Nonetheless, it would also be possible to base an unconventional course on stochastic processes on this text: presenting the many different approaches in finding a martingale associated to a peacock, and thereby introducing the necessary theory on the go, e.g., Lévy processes, random fields, harmonic functions, Fokker-Planck equations, self-similar processes, Skorokhod embedding, etc. For researchers, the book is a great opportunity to get introduced to this relatively new and fascinating branch of probability theory. For practitioners who want to create models for empirically given marginals (given, e.g., via option prices), the monograph should be a very valuable reference.