Assessing the Performance of the Block Maxima Method in Estimating Market Risk

Abstract

Extreme value theory (EVT) has emerged as one of the most important statistical disciplines for predicting the probability of unusual events from observed outliers. In the context of this theory, two approaches have been developed for selecting the extreme values of a sample and for modeling them: the block maxima method (BMM) and the peak over threshold approach (POT). This paper focuses on the block maxima method (BMM) and its ability to measure market risk. The study is conducted for a set of 15 portfolio assets that are very well diversified. To analyze the robustness of the results, we consider different block sizes. The results obtained are very conclusive. Unlike the POT method, which has been proven to be very successful in measuring market risk, the maximum block method yields very poor results. Furthermore, these results are clearly sensitive to the selected block size. The superiority of the POT method is evident. JEL classification: G17; G31