A0364
Title: Is Spearman's a robust measure of correlation for financial time series? Some evidences in the insurance sector
Authors: Roberto Baviera - Politecnico di Milano (Italy) [presenting]
Abstract: Spearman's rho is a widely used non-parametric measure of statistical dependence between two time series. Since it is rank-based, it is considered robust, as it is less sensitive to extreme values and therefore relatively resistant to outliers. It is proven that when applied to financial time series with a significant number of zeros, Spearman's rho can produce widely varying correlation estimates. In such cases, financially equivalent time series - those that should theoretically exhibit similar dependence - can record highly different Spearman's correlation. An experimental analysis is performed using datasets in the insurance sector.
