Study of Coverage Test Power for Market Risk Measures under Dependent Returns

The development and active use of financial models create a demand for a quantitative measurement of market risk. Value at Risk (VaR) remains the most popular tool for these purposes these days. Companies in banking sphere, insurance companies, investment funds and other financial enterprises use this tool worldwide. VaR allows estimating the largest amount of loss that can be incurred with a certain probability in the certain time interval. The existing variety of methods for calculating VaR requires checking validity of the models. As a prerequisite for tests to check the adequacy of VaR models, the condition of failures independence is assumed, otherwise the tests are considered invalid. On practice, failures are often dependent, but due to limited data or the researcher's neglect, testing for failures independence may be skipped. The study shows the consequences in terms of test power. The binomial test and the Christoffersen & Pelletier test were used as part of a numerical experiment for models with dependent returns. The results of this study are practical and can be applied at the corporate level in the field of market risk assessment and in related areas.