Dear all,
I have found the below contradictory.
Can someone please tell me what I am missing.
Thank you
Schweser book 4 Pg 58
Under advantages of VAR, it says ‘allows modelling the correlation of risks’
Schweser book 4 Pg 73
Under VAR based position limits – It says ‘The drawback is that the measure does not consider the correlation of the different positions. This can lead to overestimating firm VAR and misallocating capital.’
The difference is the first one your doing it the right way. You aggregate all COVAR and VAR of all the individual position to get a VaR. The second one your looking at the position limit of the VAR (the number it will fall). You add those together, which doesn’t look at correlation.
Pg 73 Basicially means, You have 3 stocks. Stock 1’s VaR is -1.5 million, stock 2 is -3 million, stock 3 is -.5 million. The sum of that is -5 million (positions). You found the maximum postiional lost, but it doesn’t use any correlation what so ever. If they aren’t 1 correlation, the total VaR is smaller than -5 million.
Thank you , so any particular reason as to why correlation is not built when arriving at total VAR using VAR by position limit.
I thought the entier purpose of ERM is to benefit from lower correlations
I think the two biggest reason to do position VaR is that it’s easy and it’s basicially worse case outcome. If you do correlation and VaR of everything you need a lot more estimates and the number of calculations go up exponentially. Add all the VaR is a piece of cake by comparison with less estimates. Also it’s likely the worse case outcome, since everything falls at the same time. The bad thing it’s like u stated in OP, capital will be misallocated due to overestimation of VaR.