Minimum capital requirements for banks in market risk: The challenge of a shift from value at risk to expected shortfall risk measure

Financial Institutions today are, more than ever before, subject to various types of risks
which shareholders, investors and regulators expect to be timely and accurately
evaluated in order to prevent financial crises. Market Risk has become one of the most
important risk sources for Financial Institutions and Basel Committee on Banking
Supervision Accords have been evolved to take into account lessons from the past. This
Thesis analyzes Expected Shortfall, the new risk measure introduced by BCBS in a recent
paper (Fundamental Review of the Trading Book, 2012) in order to capture ‘tail risk’
more effectively. Various models for Value at Risk and Expected Shortfall estimation are
presented and evaluated by backtesting techniques in two separate periods,
representing US subprime loan crisis and after crisis market conditions respectively, for
a univariate equity portfolio, using data of S&P index returns. The findings show that
parametric methods like normal and t distributions as well as the non parametric model
of historical simulation fail to produce reliable VAR and Expected Shortfall estimates for
the crisis period, as they do not respond timely in growing market volatility. Contrary to
these results, econometric models capturing volatility dynamics, represented by an
EGARCH model, seem to perform best. For this kind of models, Expected Shortfall and
VAR estimation and backtesting results indicate that these models could have acted
proactively in the beginning of the US subprime loan crisis, determining reliable and
accurate market risk limits and equivalent capital requirements. This conclusion
becomes very important for future revisions of the BCBS framework, as the vast
majority of banks adapt the simple but inefficient method of Historical Simulation for
the calculation of their Market Risk capital requirements (ΕΒΑ Report, 2017: 31,
Perignon&Smith, 2009: 367). Contrary to that, the EGARCH model fails to produce
reliable risk measure estimates in a low volatility, tranquil market condition.