hi all, I'd need a suggestion for my problem.I'd have to find a method (or a function), as: sigma(short-term) = f(sigma(long-term)), where sigma is the index standard deviation. In other words, I'd have to compute a risklevel for the short term, given the longterm one. Any reference, paper, advice very appreciated.thanx a lot.

I suggest sigma(short-term) = N(sigma(long-term), sigma1)Do some modelling to estimate sigma1, the standard deviation of sigma(short-term) over successive short-term intervals.

What sort of assets do you deal with?

hi hezza,I'm dealing with mutual funds time series.the only sure thing is the long-term volatility thatwe compute as standard deviation on historical (daily) return time serie on 5 years.

Hummm,Don't know much about your asset, sorry, but woud probably apply second order exponential filters with various time decay factors.The second order is here to remove excessive noise. Then you can apply a simple time filter saying the 1 month vol is made of 90% of 1M, 5% of one week, and the rest of instantaneous vol or something...it leads to intuitive results and can be sold to market risk depts.good luckJezza

Hi captainharlockIf you have the daily data to calculate standard deviation over 5 years, you can use this to calculate it over shorter time periods.Or does someone else provide the 5 year standard deviation for you?

hi wallace,I computed short horizon vol. on historical data.I builded a rolling window for example 3 months.So now I take only the values above the volatility-5yrsbecause I consider them as my risk (because I'm abovethe mean value that is volatility-5yrs).but now I've to say "well, my volatility on short horizonis too high compared to the volatility-5yrs" , so I'm takingtoo much risk; or I can be in the other case: "well, my volatilityis closer to the volatility-5yrs" so I'm not taking too much risk.the problem is how compute the "too much" or "not too much".thanx(.. and merry christmas to all!)

Integrating instantaneous vol over the period you require gets you the vanilla vol for that period (at least as an estimate).You are not taking too much risk with regards to the remaining tenor of your option just beause your 3 month vol is higher than the 5 year average, you are just comparing two measures of variances:- one short, for a short dated risk,- on long, for a much longer dated risk.As a result, you are mispricing the 5 year option, if your 5 year vol, once realized, is closer from the 3 months vol than from your current estimate of the 5 year vol. That's it. the risk thing cristalizes over time, not instantaneously because of the vol you choose now (except that you now have the wrong greeks...).Theoretically, you will either - loose the difference in vega now if you adjust the vol today.- loose the difference by bleeding gamma over time),This is a difficult question because you do NOT want to value a long term derivative with a short term vol.Now, in your method for estimating short/long term variance of your asset, is is important to eliminate all sticky variance methodologies. The rolling window method is one as all returns have same weight over time and fall off at the end of the window creating artificially high or low vol when dramatic event occur(as you surely know). Better to estimate a high frequency return daily by filtering higher frequency (weight is then exponential instead of flat) and then integrate that inst. variance over time in order to get your long term variance/vol.Hope if helps, good luck with it, and Merry X-mas to u

isn't your problem more a question of which basket of assets (mutual funds in your case) you want to have in your portfolio in order to minimize risk (here your measure is volatility of returns) ?