You are here

Studies in Risk Model Validation

PhD Studentship - The project will initially be following two linked themes on the topic of model validation: backtesting of risk models and the development of proxy functions

The first topic concerns the backtesting of risk models

The Value-at-Risk (VaR) risk measure is a central concept in the Basel regulatory capital framework for banks and Solvency 2 for insurers.

However, there has been a lot of interest in the alternative expected shortfall (a.k.a conditional VaR or tail VaR) risk measure, which is coherent. The Basel Committee is proposing that banks move to the expected shortfall measure for the trading book (see Fundamental Review of the Trading Book 2012). The student will review the requirements for a risk measure to be elicitable and then explore how the underlying mathematical ideas can be adapted for use in a banking setting.

Is elicitability really relevant in a regulatory setting? Is there a better alternative? Does a move to elicitable risk measures constrain firms' ability to game current backtesting regimes that are based on VaR exception counts? What is an acceptable score in a backtesting regime based on an elicitable risk measure? How do we decide whether to accept/reject a submitted set of backtesting results? How easily can the decision be defended against the complaints of the regulated?

Are there risk measures that are both coherent and elicitable and could these be used in model validation in future?

The second topic concerns the development of proxy functions

Proxy functions provide an analytical approximation to the value of a set of liabilities using current information without the need for computationally intensive simulations. Proxy functions are felt to be potentially very useful in determining SCR's under Solvency II and many other problems involving a requirement to project insurers' balance sheets. This part of the model validation project will consider a number of specific situations and seek to validate the use of specific proxy functions in these situations. A key issue will be the development of computationally efficient algorithms.

ARC research project: Studies in Risk Model Validation
ARC scholar: Hsiao-Yen Lok
University: Heriot Watt University
Period of research: January 2014 – July 2017
Outputs:

You might also like

Contact Details

If you have any questions or wish to discuss any aspect of research carried out through the Actuarial Research Centre (ARC), please contact the Research and Knowledge Team.

arc@actuaries.org.uk