Abstract Solvable Markov Processes

This thesis introduces a new method of constructing analytically tractable (solvable) one-dimensional Markov processes- processes for which the transition probability density can be computed explicitly. For this purpose we introduce the concept of stochastic transformation as a way to map solvable diffusion process into a solvable driftless process. Stochastic transformations consist of an absolutely continuous measure change and a diffeomorphism. Our main theorem characterizes all stochastic transformations and gives a remarkably simple algorithm to construct all the stochastic transformations for a given process. We study in detail properties of these transformations and the boundary behavior of transformed processes. As examples we show how one can obtain the well known and widely used martingale models (quadratic volatility, CEV processes) as well as obtain new important classes (Ornstein-Uhlenbeck, CIR and Jacobi families) of solvable driftless processes. Inspired by these ideas, in the next chapter we classify all solvable one-dimensional driftless diffusions, for which the transition probability function can be computed as an integral over hypergeometric functions. In the last chapter we give several examples of how one could use the above processes for financial modelling. First we prove the result that lattice approximations to Ornstein-Uhlenbeck and CIR processes, given by Charlier and Meixner processes, are affine and compute explicitly the generating and character