Inferring jump dynamics from weekly options: A non-parametric method

Abstract

With the increasing demand for short-term information extraction, this paper explores non-parametric methods for proxying risk-neutral jumps. Empirical evidence from weekly options shows that among the approximation measures, the non-parametric jump derived from the linear term in the quadratic relationship between cubic returns and time-to-maturity has a significantly positive slope coefficient with one-day-ahead returns. In out-of-sample tests, its predictive power remains robust, showing the best performance with a short training window. Additionally, the predictive power of the jump derived from the slope of the cubic return for the two adjacent shortest times to maturity improves when both jump and risk-neutral variance are integrated into a single model.