Normal vs lognormal VaR, Tail distribution


New Member

1. I've recently watched a video of normal vs lognormal VaR where David explained that one of the most prominent advantages of lognormal VaR is that we cannot end with a VaR larger than the portfolio value.

I have a practical question then though: if I were to calculate a VaR on my portfolio - do I in reality have a discretion over which distribution to use? Is there a common practice / preference among pracitioners over which one to use?

2. What is a common approach with regards to modeling the tail nowadays? Is it really the beta/GEV distributions where the parameters are tailored so to match more or less the historical shape of a tail?

Appreciate my questions might be slightly generic.

Many thanks