Value at Risk (VaR). Empirical parametric models and evaluation

• Main Author: Ιωαννίδης, Μιχαήλ
• Other Authors: Βενέτης, Ιωάννης
• Format: Thesis
• Language: English
• Published: 2018
• Subjects: GARCH; Variance-covariance; Historical simulation; Kupiec’s unconditional coverage test; Christoffersen’s independence test; Christoffersen’s joint test; Mixed Kupiec test; Μήτρα διακύμανσης-συνδιακύμανσης; Ιστορική προσομοίωση; Έλεγχος του Kupiec; Έλεγχος ανεξαρτησίας του Christoffersen; Μεικτός έλεγχος του Christoffersen; Independence test by Haas; Έλεγχος ανεξαρτησίας του Haas; Μεικτός έλεγχος του Kupiec; 332.6

We compare the performance of two parametric methods with one non parametric for day-to-day Value-at-Risk models for 63 stocks traded in NYSE or NASDAQ for three different sample sizes and two significance levels. We find that Historical Simulation is more accurate, but also more dependent on its past VaR violations than GARCH and Variance Covariance model.