150 likes | 167 Views
Friends or Foes: A Story of Value at Risk and Expected Tail Loss. By Saša Žiković. Discussant: Iftekhar Hasan Rensselaer Polytechnic Institute. Objective. This study proposes Using advanced Value at Risk (VaR) calculation techniques to produce superior Expected tail loss (ETL) forecasts
E N D
Friends or Foes: A Story of Value at Risk and Expected Tail Loss By Saša Žiković Discussant: Iftekhar Hasan Rensselaer Polytechnic Institute
Objective • This study proposes • Using advanced Value at Risk (VaR) calculation techniques to produce superior Expected tail loss (ETL) forecasts • a new loss function that relates tail losses to ETL forecasts • This study empirically tests • VaR and ETL models using US and European stock indexes
Summary • Extensive literature review on VAR and ETL estimation • Critiques of shortcomings associated with the existing methodologies • Develop a new semi-approach to estimating ETL
Key findings and conclusions • Bootstrapped Hybrid Historical Simulation ETL approach is the best performer • The Back-testing are conducted to validate the findings or claims
Comments • Motivation • This study is not well motivated • The fact that very few studies using ETL does not necessarily mean it is important • This study is an empirical one in nature, and, therefore, needs to point out its empirical importance • Using an illustrative example could help to explain why tail loss is important in this setting of risk management
Comments (cont’d) • Contribution • What is the value added to existing study? • A clear cut perspective is necessary to distinguish what is the contribution (proposition, model, measure, and findings) of this paper and what has been done by existing literature • What is the relevance to existing study? • This paper needs also to point out the relevance of what is conducted in this paper to the exiting literature • An improvement of exiting approach? • An new approach? • New findings?
Comments (cont’d) • Literature • Literature review here should not be a comprehensive survey • Literature review should • Briefly portray what has been done in this field • Find out the gap of knowledge • Clarify why this paper is important and necessary • Possibly point out how this study is proceeded
Comments (cont’d) Estimation procedure • Briefly cite other methods, give key points and refer them to their papers. • Elaborate on your own methods as detailed as possible, point out why they are better. • Discuss alternative procedures even if it is for robustness purposes.
Comments (cont’d) • Findings • The findings are ad hoc • What is the context for using this approach? • What is the implication for decision making for practitioners (e.g., bankers, fund managers, etc) ?
Writing style • Verbosity • Literature should focus on • Shortcoming of VaR • Importance of ETL • The connection between VaR and ETL • For example, if you the author include any discussion on the lack of coherence of VaR and the Coherence of Cvar (more known as expected shortfall or expected tail loss (ETL) to motivate the goal of estimating ETL, simply note that the marginal of risk with a VaR constraint is, in fact, coherent and cite a paper (Eisenberg, Journal of Risk, July, 2007).
Writing style Clarification • For example, “changes in the efficiency of the market” (bottom p.13) Consistency • For example, what are VCV, HS250 or RiskMetrics ? (in the back-testing section)
Caution • In section 4 (pages10 and 11) the paper provides some brief background on the use of extreme value theory (EVT) in other fields such as hydrology and make your case for not using it to estimate ETL. In section 4 half a page from the end of that section (p. 12) which has 6 ½ pages, you propose using an approach developed in your earlier work that uses “hybrid historical simulation (HHS)” as a “basis for developing a semi-parametric approach to estimation of ETL.”
Caution • You propose a “boostrapped HHS ETL model that starts by standardizing the tail losses in excess of HHS VaR by the latest GARCH volatility update.” “Since these standardized tail losses are now IID they are suitable for boostrapping.” However, (p.12) you state that you use Embrechts approach. So it is not clear yet what is the key contribution. • The opportunity for the paper to make a contribution is in the testing of the Embrechts approach by fitting your ETL estimates and then testing them out of sample.
Caution What is the total benefit (loss) in using (not using) this new method. What would be the optimum capital to hold under each scenario? More transparent benchmarking may be warranted.
Future direction • Clearly, you have the technical skill and the understanding of the subject matter. • You need to say how others can follow this method – what context, what potential data sets and time period or countries. This would create others to follow your methods citing this approach.