If I am doing a 99th percentile VaR over 1 month on my balance sheet for say equity risk does that mean:{surplus at t=0} - {surplus in 99th percentile at t=1} (possibly allowing for discounting from t=1 to t=0)Is that right - or does someone have another definition? What are views on discounting from t=1 to t=0? My view is that if the requirement is VaR over 1 month then discounting is not correct, but if the requirement is in 1 month then it is correct.

thanks a lot for the quick reply; and good question, that is helpfulinsurance companies have to a do VaR on one year, which makes the discounting more of a question (though perhaps not a lot in currently low yield context)I have heard another definition of VaR as:{{surplus at t=1 in real world best estimate} - {surplus in 99th percentile at t=1}}/{1+some discount rate}this definition makes no sense to me as it has no practical meaning and increases capital on assets with high expected return for no real reason. Further, if both the best estimate and 99th percentile return are positive - you would still need to hold capitalI wondered if there is a paper anywhere with a formal definition?Many thanks for any help

You seem confused at first glance.If by balance sheet you mean the classic tautology assets-liabilities=equity, you need to understand that items on a balance sheet represent levels of a stock, at a given point in time. Whatever you name as "surplus" can only imply a flow over a period between 2 points in time. A 1 month 99% VaR of a portfolio of X or x%, depending on if it denotes value or returns, is just a forward looking estimate that in one month from now there is only 1% chance the porfolio to worth (in market to market terms) less than X or yield less than x%, if the portfolio remains as is today for the entire month, something that doesnt happen too often.The next vague point is the value upon which you estimate VaR. The right approach I guess would be estimate VaR upon your assets. So if your assets VaR is estimated as X, then you need to subtract from X an estimate about your level of liabilities at the end of the month in order to reach to VaR upon your equity. Your end of month liabilities would be current liabilities+new liabilities-redemptions+accrued interest for the period. HTH

QuoteOriginally posted by: outrunThat's correct. Discounting is typically an effect too small to be relevant in the context of the VAR size movements (and the precision of that values).However a good thought is: what VAR value would you like to see in case you've invested in a zero volatility strategy?I am against discounting, not because of the negligible effect, but because in estimating VaR you try to have a view on what the future position, structured as is today, will be worthing at the end of the period, not what the future position worths now. If you need the later piece of info, MtM approach works the best.

thanks for this. What if the purpose of the VaR calculation is calculation of capital which you are required by regulation to set aside today - is discounting the right approach?If the regulation states VaR is to be calculated IN 1 years time, I think discounting is possible. If it states OVER 1 year I think it is not possible as the capital could be required at any point over the next year (including tomorrow).