Explanation:
When backtesting VaR, a 99% confidence level implies very few exceptions (only 1% of the time, or roughly 2.5 days per 250-trading-day year). Because the expected number of exceptions is so low, it becomes statistically difficult to definitively differentiate between a good model and a flawed model (a problem known as low statistical power).
By lowering the confidence level to 95%, the expected number of exceptions increases (about 12.5 days per year). This larger sample size of tail events increases the statistical power of the backtest. Therefore, the accuracy of the test and the reliability of accepting or rejecting the VaR model are significantly greater at a 95% confidence level compared to 99%.
Ultimate access to all questions.
Q.53 While carrying out backtesting of a leading bank's VaR model, you have made the following findings: the bank is currently calculating the 1-day VaR at a confidence level of 99%. However, based on your findings you suggest changing the confidence level from 99% to 95%. Which of the following statements would justify your stance?
A
While conducting backtesting with a 95% confidence interval, the probability of committing a Type 1 error is small as compared to the probability of Type 1 error when backtesting with 95% and 99% VaR models.
B
The accuracy of the VaR model and the basis of accepting/rejecting the model have a greater reliability at a 95% confidence level VaR as compared to at a 99% confidence level.
C
While conducting backtesting with a 95% confidence interval, the probability of rejecting the VaR models at a 95% confidence level is equal to that at a 99% confidence level.
D
There are fewer chances of a 95% VaR model being rejected based on backtesting as compared to a 99% VaR model.
No comments yet.