Title:On stationary Markov processes with polynomial conditional moments

Abstract:We study class of stationary Markov processes with marginal distributions identifiable by moments. We assume that this class has additional property that every conditional moment of degree say $m$ is a polynomial of degree at most $m.$ We show that then under some additional, natural technical assumption here exist a family of orthogonal polynomial martingales.
We single out also two important subclasses of the considered class of Markov processes. Namely the class of harnesses that we characterize completely within the considered class of Markov processes. The second subclass constitute of the process that we call process with independent regression property that is process that so to say generalizations of ordinary Ornstein-Uhlenbeck processes. Generalization in the sense that the marginal distributions are not necessarily Gaussian but is infinitely divisible while $X_{t}-E(X_{t}|X_{s})$ for $t>s$ is independent of $X_{s}.$ We show that these process are not time scale transforms of Lévy processes.
As a corollary of our considerations on harnesses we characterize the so called quadratic harnesses within the class of Markov processes with polynomial conditional moment property. We show that it is relatively poor i.e. there exist only one parameter $q\in \lbrack -1,1]$ that defines stationary distribution. The case $q=1$ leads to Ornstein--Uhlenbeck (OU) process, $q\in (-1,1)$ leads to the so called $q-$OU process, while $q=-1$ leads to simple symmetric two point state Markov chain.