On The Distribution Tail Of Stochastic Differential Equations: The One-Dimensional Case (1601.03858v4)

Abstract: This paper considers a general one-dimensional stochastic differential equation (SDE). A particular attention is given to the SDEs that may be transformed (via Ito's formula) into:$$d X_t = ( \bar{B} (X_t) - b X_t) d t + \sqrt{X_t} d W_t, ~~~X_0 > 0,$$where $ \bar{B}(y)/ y \to 0$. It is shown that the MGF of $X_t$ explodes at a critical moment $\mu^\ast_t$ which is independent of $\bar{B}$. Furthermore, this MGF is given as a sum of the MGF of a Cox-Ingersoll-Ross process plus an extra term which is given by a nonlinear partial differential equation (PDE) on $\partial_t$ and $\partial_x$. The existence and the uniqueness of the solution of the nonlinear PDE is then proved using the inverse function theorem in a Banach space that will be defined in the paper. As an application, the mean reverting equation $$d V_t = ( a - b V_t) d t + \sigma V^p_t d W_t, ~~~V_0 = v_0 > 0,$$is extensively studied where some sharp asymptotic expansions of its MGF as well as its complementary cumulative distribution (CCDF) are derived.