A1083
Title: Comparing value at risk and expected shortfall estimation using LSTM and EGARCH family models
Authors: Shujie Li - Paderborn University (Germany) [presenting]
Abstract: The forecasting performance of value at risk and expected shortfall is examined across several models from the EGARCH family, including the newly introduced modulus log-GARCH, a modified EGARCH, and their long memory extensions, as well as a recurrent neural network based on long short term memory architecture. To assess model performance, eight stock indices from diverse international markets are analyzed. The models are evaluated using three different back-testing approaches and a model selection criterion, namely the weighted absolute deviation. The results indicate that the selected indices exhibit heavy tails and asymmetry. Therefore, in general, models estimated with skewed distributions perform better than their symmetric counterparts. In most cases, the long short term memory model is selected as the top-performing model. Nevertheless, several models from the EGARCH family remain strong competitors, especially for asymmetric distributions, and might be preferred for certain indices.
