Evaluation of the Predictive Ability of VaR Models during Different Market Conditions.

Liew, KeiYan (2014) Evaluation of the Predictive Ability of VaR Models during Different Market Conditions. [Dissertation (University of Nottingham only)] (Unpublished)

[img] PDF - Registered users only - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (1MB)


Risk management methods in finance have put a lot of weight on the

Value-at-Risk, making it the most popular risk measurement tools. The

main purpose of VaR models is to capture future market risks accurately.

Thus it is important to gauge the predictive ability of VaR estimates.

Consequently, backtesting methods have been developed in order to

systematically compare VaR estimates to actual market losses and profits.

It is important that the appropriate backtest techniques are used.

Subsequently a model is accepted or rejected based on these tests. This

thesis consists mainly of a contribution to empirical studies of VaR. The

empirical research of this thesis focuses on the performance of the FTSE

index during years leading to the 2008 Global Financial Crisis till after the

crisis. The main aim of this study is to evaluate the predictive ability of

VaR models in foreseeing the 2008 crisis which affected the UK. VaR

forecasts are estimated by various VaR models in three periods of

different market conditions. The secondary aim attempts to choose the

most reliable backtest. A third and sub-objective, is to provide a

benchmark study on the performance of VaR models when time horizon is

varied. This may provide valuable information on the rigidity or flexibility

of a model and thus its role in other topics of risk management e.g crisis

management. Tests of unconditional and conditional coverage are applied

together with tests of independence. Empirical Coverage Probability and

Basel Traffic Light test are also employed. This thesis is a univariate

analysis focused in the UK. 1-day, 5-day and 10-day VaR estimates for a

one year time period are used in the backtesting process. Results from

the backtesting shed light on potential problems within the VaR system.

As time horizons increase, severe underestimation and dependence of risk

are observed. Furthermore, unstable market conditions cause problems in

backtesting evaluation since VaR models are commonly recognized to

perform well only under normal market conditions.

Item Type: Dissertation (University of Nottingham only)
Depositing User: EP, Services
Date Deposited: 12 Nov 2014 09:59
Last Modified: 19 Oct 2017 13:58
URI: https://eprints.nottingham.ac.uk/id/eprint/27448

Actions (Archive Staff Only)

Edit View Edit View