wilmott.com

Hi,Suppose I want to calculate VaR using historical returns and prices with a portfolio that has options.What volatility should I use for the option valuations?? Should I use the implied volatility from the same day that the foward/spot and interest rate were taken??ie..my time series I am sampling from to get my historic data for an option on spot are: SpotPrice InterestRate ImpliedVolatilityday1 day2day3.........day250then, I choose a random day, and then use the (SpotPrice, InterestRate, ImpliedVolatility) to calculate on that day the value of the option, then (assuming my VaR is over one day) I take the next day's (SpotPrice, InterestRate, ImpliedVolatility) to work out the option price on the final day.Is this correct? Or should the volatility come from somewhere else?Cheers,Tony

It is very confusing whatever you are doing. If the portfolio existed historically, then why don't you have option prices?If the portfolio exists only today (and in the future), and you are using history just to sample the underlying,then maybe it makes more sense to take today's implied vols, or to sample the implied vols, or to do something else.

Hi AlanThanks for the reply and pointing out what I wrote was confusing, I think it was plain wrong. So here's my attempt to write it clearly.I want to calculate VaR using historic sampling. For swaps/forwards/spot we create the return for each underlying from day t to day t+1 in the past over the time period we have the historic data for, for swaps/forwards/spot the change in value over 1 day can be calculated using a randomly drawn return from the timeseries. However, for an option, my understanding is that we need to calculate the option price today and the option price tomorrow and take the difference.Calculating the option price today is no problems as we have all the input data, however for the calculation of the price tomorrow we can use a randomly drawn return to bump the underlying, but I am not sure what volatility to use, would we bump the volatility by the same amount that was experienced on the same day that the randomly drawn return was taken from? ie something likevol_change = (vol_t - vol_{t-1}) / vol_{t-1} this doesn?t seem right to me, as we don?t really think of returns on volatility (at least I don't)?Cheers,Tony

I see. Here is what I would do. For each underlying, you can get a return, as you explained, which gives youthe day t+1 underlying price. To convert this into a day t+1 option price, you need a day t+1 implied vol.Say the underlying is IBM. To get that implied vol., in the same spirit of what you are doing, I would make a random draw from all the implied-vols associated to options on IBM that are present in your data set.