Excel is enough.

Historical simulation means you assume the past N day return series is drawn from an unknown distribution. Say you have 100 past returns. You then try to find the 5% quantile of this distribution(assume you want 95% VaR). What you need to do is just sort this 100 returns in the ascending order and find the 5th number. This means there is only 5 numbers out of 100 smaller or equal to this number. So this is 95% VaR.

Monte Carlo simulation is always combined with GARCH model. It assumes that you know the distribution of the return (e.g. normal, or t) you use GARCH model to estimate the parameters from the historical data and then use the GARCH model to simulate a path of volatility so that you can get the corresponding VaR's path. Simulate enough paths (10000) average them and you get the VaR. This method is always used to predict VaR of a long time horizon (e.g. you have daily return but you want to predict 10 day VaR)

best,

谢谢MZ同志，但是蒙特卡罗模拟法还是不是很懂，因为我未接触过GRACH模型，现在我只会用normsinv(rand(),0,1)模拟出10000个数据，请问下一步怎样呢？

in that case, I think it is unnecessary to use MC method. Since you know the return is Normal, then you can calculate the 95% or 99% left tail quantile directly without simulation.

If you really want to do it via MC, you can simulate 10000 path and sort them in the ascending order and pick the 10th or 50th number to be the 99% or 95% VaR.

Best,

thanks very much!!

支持一下，留个脚印