Hello, a quick question is how do we know when to use absolute VaR vs relative VaR? The answer ended up using relative VaR because I supposed enough information was not given anyway. But any additional insight would be helpful. Also why was the lognormal VaR not used?
Hi @dbansal Re normal/lognormal distinction: the question is flawed for not specifying normal/lognormal VaR. It's from 2011, we gave feedback on this for years such that GARP's modern questions know to specify (see their 2017 practice paper for evidences of better specification). This question says "assuming returns of BNA are normally distributed" but astute readers will notice that is insufficient. The question would need to say something like either "arithmetic returns are normally distributed" (aka, normal VaR) or "geometric returns are normally distributed" (aka, lognormal VaR) but GARP should be well enough aware to go beyond even this distinction and simply request a "normal VaR" or a "lognormal VaR."
Re: relative/absolute: We've also given a lot of feedback on this, and I expect current FRM to be clear on what it wants. You are right that 2011.P1.13 is sketchy because you can't really do anything except assume µ=0, but my position is that the question is imprecise here too. The old questions failed on this often, but recent questions are aware.
These two issues are exactly why we (BT) are working on an open-source glossary that will be shared with GARP and the EPPs: the anthology approach leaves too many generic terms with unclear, or multiple, definitions if they are not clarified. These two are good examples. I hope that helps!