Editorial Director
Managing Editor
 
PRESENTATION ARCHIVES
  N°157  
Issue Q1 2019  
The accuracy of asymmetric GARCH model estimation  
Amélie Charles
Olivier Darné
 
This paper reviews eight software packages when estimating asymmetric GARCH models (from their default option). We consider the numerical consistency of GJR-GARCH, TGARCH, EGARCH and APARCH estimations with Normal and Student distributions as well as out-of-sample forecasting accuracy, using the model confidence set procedure. We show that results are clearly software-dependent for both asymmetric volatility models, especially for the t-ratios. The out-of-sample forecast results show that the differences in estimating symmetric and asymmetric GARCH models imply slight differences in terms of forecast accuracy, not statistically significant, except in few cases from the QLIKE loss function. Further, the results indicate that the different specifications of the asymmetric GARCH-type models used by the different packages appear to have no significant effect on their forecast accuracy.
Abstract

   
EGARCH ; GJR-GARCH ; TARCH ; APARCH ; Accuracy ; Forecasting ; Software ; Keywords
C22 ; JEL classification
Order form