MPRA Paper No. 37023
24 Pages Posted: 9 Mar 2012
Date Written: March 2012
Extreme Value Theory is increasingly used in the modelling of financial time series. The non-normality of stock returns leads to the search for alternative distributions that allows skewness and leptokurtic behavior. One of the most used distributions is the Pareto Distribution because it allows non-normal behaviour, which requires the estimation of a tail index.
This paper provides a new method for estimating the tail index. We propose an automatic procedure based on the computation of successive nor- mality tests over the whole of the distribution in order to estimate a Gaussian Distribution for the central returns and two Pareto distributions for the tails. We find that the method proposed is an automatic procedure that can be computed without need of an external agent to take the decision, so it is clearly objective.
Keywords: Tail Index, Hill estimator, Normality Test
JEL Classification: C10, C15, G19
Suggested Citation: Suggested Citation
Gimeno, Ricardo and Gonzalez, Clara I., An Automatic Procedure for the Estimation of the Tail Index (March 2012). MPRA Paper No. 37023. Available at SSRN: https://ssrn.com/abstract=2018824 or http://dx.doi.org/10.2139/ssrn.2018824