Medición del riesgo de la cola en el mercado del petróleo mexicano aplicando la teoría de valores extremos condicional

Abstract

This paper applies the extreme values theory to the conditional distribution of standardized residuals from the specifications GARCH, EGARCH and TGARCH, and proposes dynamic risk measures to estimate VaR and expected shortfall of long and short positions of the Mexican Blend crude oil from January 4, 1989 to December 31, 2013. The results of backtesting procedure show that the models based on the conditional extreme value theory and filtered historical simulation yield more accurate estimates of conditional VaR at all confidence levels although their performance is lowered significantly for the conditional expected shortfall prediction. At 99.5% and 99.9% confidence levels, the empirical findings reveal that the government is prone to experience a higher risk than the consumers of Mexican crude oil at the international market because the inferior tail of empirical distribution is more stable and heavier than the superior tail. (English) [ABSTRACT FROM AUTHOR]