|
|
 |
| |
|
|
|
Integrating Composite Models for Enhanced Risk Assessment of the South African Industrial Index (J520) |
|
|
|
PP: 731-747 |
|
|
doi:10.18576/jsap/140616
|
|
|
|
Author(s) |
|
|
|
Sandile C. Shongwe,
Zander Greyling,
Frans F. Koning,
|
|
|
|
Abstract |
|
|
| In this paper, we examine whether composite models provide superior performance to single-distribution
approaches in modeling financial return of the South African Industrial Index (J520). The composite models are developed by piecing together two distributions at a threshold value, where the portion from the left tail to threshold value (captures the frequent but low-to-moderate severity events) is called the ‘head distribution’ while from the threshold to the right tail (captures the rare but high severity events) is called ‘tail distribution’. Thus, using 16 different single distributions, we construct and evaluate a total of 256 composite models, i.e. (16 for the ‘head component’) × (16 for the ‘tail component’). Model performance is assessed by using standard goodness-of-fit criteria and by estimating key risk measures at the 95%, 99%, and 99.5% confidence levels. Loss and gain returns are analyzed separately, with the top 20 models (out of a total of 256) reported in terms of both model fit and risk estimation accuracy. By systematically comparing composite and single- distribution frameworks, the study addresses the practical question of whether greater model flexibility translates into improved risk assessment for real-world financial data. The findings contribute to the growing literature on composite
modeling and its applications in actuarial science and risk management. |
|
|
|
|
|
|
|
