CMStatistics 2017: Start Registration
View Submission - CFE
Title: Hypergeometric functionals and kernel regression in risk neutral density estimation Authors:  Antonio Santos - University of Coimbra (Portugal)
Ana Monteiro - University of Coimbra (Portugal) [presenting]
Abstract: The analysis of financial risk through the information contained in options prices has been for long time considered important, and one way to extract such information is through the indirect estimation of the risk-neutral densities (RND). We compare different methods to retrieve information from estimated RND. There are several approaches for the estimation: parametric, semi-nonparametric and nonparametric methods, in particular, we consider the use of hypergeometric density functionals, and less structured methods like the ones based on kernel regression density estimation. We address the computational challenges associated with the use of hypergeometric functions, and the need of different amounts of data. The comparisons are made through simulated data and also from intraday market data, which to our knowledge is a novel approach. Due to the fact that the RND is not observable, a simulation analysis will attest that the methods used are able to capture the ``true'' density (in case of simulated data), and afterwards both methods are applied to real data sets. The root mean integrated error (RMISE) criterion is used to assess the quality of the estimation.