Expected Shortfall fixes some of the disadvantages that VaR has. VaR says how much loss can be expected at a certain likelihood. E.S. tells you more about the shape of the bell curve and what happens beyond the VaR level.
Could you pls help explain that does VaR provide a maximum or a minimum value we can lose for a given confidence level ? Or does it depends on the type of distribution i.e profit distribution or loss distribution? please help clarifying this as I'm little confused. thanks
Hello @mamta0508. Focusing exclusively on the loss distribution, the use of "maximum" vs "minimum" is respectively linked to the whether you use "level of confidence" vs "level of significance" to express your VaR. If your 95% confidence VaR is 10 mn, you will be 95% confident that your actual loss (or negated profit) will not exceed 10 million (i.e. a maximum level). Alongside, you can say that 5% of the time, your loss will turn out to be more than 10 million (i.e. a minimum level).
@@mamta0508Just to be clear, 95% VaR = 10mn gives just one number. 95% of the time (e.g. 19 times out of 20) the Clean P&L (essentially excluding factors not in Var, e.g. income from new deals and time effects) should be either a profit (of any size) or a loss in the range 0-10mn. It doesn’t say how bad the loss might be, in the 5% of the time that VaR is exceeded! With HistSim VaR, you can monitor the tails to some extent by logging VaR at other confidence intervals, e.g. 97.5% and 99%. If the distribution is normal then the ratio of 99% VaR / 97.5% VaR should be about 1.18, so if you observe a bigger ratio than that then you might have “fat tails”.
Great insight.
Expected Shortfall fixes some of the disadvantages that VaR has. VaR says how much loss can be expected at a certain likelihood. E.S. tells you more about the shape of the bell curve and what happens beyond the VaR level.
Isn't the left tail is for the loss distribution and right tail for the profit distribution?
Could you pls help explain that does VaR provide a maximum or a minimum value we can lose for a given confidence level ? Or does it depends on the type of distribution i.e profit distribution or loss distribution? please help clarifying this as I'm little confused. thanks
Hello @mamta0508. Focusing exclusively on the loss distribution, the use of "maximum" vs "minimum" is respectively linked to the whether you use "level of confidence" vs "level of significance" to express your VaR. If your 95% confidence VaR is 10 mn, you will be 95% confident that your actual loss (or negated profit) will not exceed 10 million (i.e. a maximum level). Alongside, you can say that 5% of the time, your loss will turn out to be more than 10 million (i.e. a minimum level).
@@finRGB thank you very much for clearing this doubt!
@@mamta0508Just to be clear, 95% VaR = 10mn gives just one number. 95% of the time (e.g. 19 times out of 20) the Clean P&L (essentially excluding factors not in Var, e.g. income from new deals and time effects) should be either a profit (of any size) or a loss in the range 0-10mn. It doesn’t say how bad the loss might be, in the 5% of the time that VaR is exceeded! With HistSim VaR, you can monitor the tails to some extent by logging VaR at other confidence intervals, e.g. 97.5% and 99%. If the distribution is normal then the ratio of 99% VaR / 97.5% VaR should be about 1.18, so if you observe a bigger ratio than that then you might have “fat tails”.
Great video as always ! I just have a quick question, what's the difference between the usual VaR and Credit VaR ?
Thank you for the appreciation, Haythem. Will do a short video on Credit VaR.
@@finRGB Thank you so much, can't wait to see it !