Market Risk Management: The Applicability and Accuracy of Value-at-Risk Models in Financial Institutions
(2011)Department of Business Administration
- Abstract
- 2
Abstract
The paper focuses on evaluating the most performant Value-at-Risk models from the perspective of financial institutions. A number of 18 VaR methodologies were used for this purpose, comprising of parametric and non-parametric methods, some of which include time-varying volatilities estimated by means of GARCH and asymmetric GARCH models.
All methods were applied on real P/L data extracted from four commercial banks. Furthermore, we use two backtesting frameworks to validate the results: the Kupiec test (1995) and the newly developed Risk Map (2011). The use of the latter method allowed us to account not only for the number of tail losses but also for the magnitude of these exceptions, demonstrating once again its reliability and... (More) - 2
Abstract
The paper focuses on evaluating the most performant Value-at-Risk models from the perspective of financial institutions. A number of 18 VaR methodologies were used for this purpose, comprising of parametric and non-parametric methods, some of which include time-varying volatilities estimated by means of GARCH and asymmetric GARCH models.
All methods were applied on real P/L data extracted from four commercial banks. Furthermore, we use two backtesting frameworks to validate the results: the Kupiec test (1995) and the newly developed Risk Map (2011). The use of the latter method allowed us to account not only for the number of tail losses but also for the magnitude of these exceptions, demonstrating once again its reliability and more importantly its simplicity in application.
Therefore, the study aims to add to the relatively scarce current literature that investigates the use of VaR models in financial institutions, using real bank data.
The results showed that that models working under the assumption of normality are not the most performant when calculating VaR, models under the assumption of the t-distribution providing more performant VaR estimates.
Parametric models that use GARCH forecasted volatilities have outperformed the other methods, especially the ones using a t-distribution for both the data series and the innovations. These models both passed the backtests and produced some of the lowest average VaRs, which is important for financial institutions in terms of minimizing their capital requirements. Although in some situations the asymmetric GARCH models showed a better performance than the simple GARCH model, the increased efficiency is not significantly superior.
As for non-parametric models, the basic historical simulation passed the backtest for all institutions, even though it is a simple measure and yields a relatively flat structure of the VaR forecasts over time. In addition, the newly developed HS-VIX model also worked well for all the data series, in providing accurate VaR estimates throughout the backtesting window. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/1981848
- author
- Birtoiu, Alina and Dragu, Florin George
- supervisor
- organization
- year
- 2011
- type
- H1 - Master's Degree (One Year)
- subject
- keywords
- Management of enterprises, Företagsledning, management
- language
- Swedish
- id
- 1981848
- date added to LUP
- 2011-06-01 00:00:00
- date last changed
- 2012-04-02 18:51:46
@misc{1981848,
abstract = {{2
Abstract
The paper focuses on evaluating the most performant Value-at-Risk models from the perspective of financial institutions. A number of 18 VaR methodologies were used for this purpose, comprising of parametric and non-parametric methods, some of which include time-varying volatilities estimated by means of GARCH and asymmetric GARCH models.
All methods were applied on real P/L data extracted from four commercial banks. Furthermore, we use two backtesting frameworks to validate the results: the Kupiec test (1995) and the newly developed Risk Map (2011). The use of the latter method allowed us to account not only for the number of tail losses but also for the magnitude of these exceptions, demonstrating once again its reliability and more importantly its simplicity in application.
Therefore, the study aims to add to the relatively scarce current literature that investigates the use of VaR models in financial institutions, using real bank data.
The results showed that that models working under the assumption of normality are not the most performant when calculating VaR, models under the assumption of the t-distribution providing more performant VaR estimates.
Parametric models that use GARCH forecasted volatilities have outperformed the other methods, especially the ones using a t-distribution for both the data series and the innovations. These models both passed the backtests and produced some of the lowest average VaRs, which is important for financial institutions in terms of minimizing their capital requirements. Although in some situations the asymmetric GARCH models showed a better performance than the simple GARCH model, the increased efficiency is not significantly superior.
As for non-parametric models, the basic historical simulation passed the backtest for all institutions, even though it is a simple measure and yields a relatively flat structure of the VaR forecasts over time. In addition, the newly developed HS-VIX model also worked well for all the data series, in providing accurate VaR estimates throughout the backtesting window.}},
author = {{Birtoiu, Alina and Dragu, Florin George}},
language = {{swe}},
note = {{Student Paper}},
title = {{Market Risk Management: The Applicability and Accuracy of Value-at-Risk Models in Financial Institutions}},
year = {{2011}},
}