This paper investigates the effects of using temporal aggregation rules in the evaluation of the maximum portfolio loss1. In particular, we propose and compare different time aggregation rules for VaR models. We implement time-scale transformations for: (i) a EWMA model with Student's t conditional distributions, (ii) a stable sub-Gaussian model, (iii) a stable asymmetric model. All models are subjected to backtest on out-of-sample data in order to assess their forecasting power and to show how these aggregation rules perform in practice.
|Title of host publication||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Editors||Marian Bubak, Geert Dick van Albada, Peter M. A. Sloot, Jack J. Dongarra|
|Number of pages||8|
|State||Published - 2004|
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|