A highly sensitive mean-reverting process in finance and the Euler-Maruyama approximations (Q947594)

The paper studies a class of one dimensional stochastic differential equations of the general form \(dR(t)=\lambda \, (\mu - R(t)) \, dt + \sigma \, R^{\gamma}(t) \, dW(t)\), where \(W(t)\) is a scalar Wiener process, for \(\gamma >1\), positive parameters and \(R(0)>0\), which model mean reverting Itô processes. These are generalizations of popular models in financial mathematics, e.g. the Cox-Ingersol-Ross model or the Vasisec model for interest rates etc, which seem to be supported by data analysis. This model is not covered by the standard results on stochastic differential equations (SDEs) available in the literature, which assume linear growth conditions for the coefficients of the SDE. The paper develops a theory for the qualitative properties of the solution of this equation, proving at first existence and uniqueness of global positive solutions, boundedness of the first and second moments of the solution, stochastic boundedness results which guarantee that the solution stays in a belt area with a large probability, as well as pathwise estimates for the growth of solutions. Then the application of numerical methods of the Euler-Maruyama type to this class of equations is studied, by providing detailed error estimates and convergence results for the numerical method. Finally, the theoretical results concerning the Euler-Maruyama method are applied to the problem of the valuation of a variety of financial contracts when the underlying asset follows a stochastic process modelled by the SDE in question, by proving the convergence of the price predicted by the disretized SDE to the price predicted by the full model.