wilmott.com

A very practical problem:Our system gets options prices from Reuters every day. Then we calculate theimplied vol using the options price.If the option is liquidly traded, then the option?s price is updated andthe implied vol is correct. Actually sometimes options are not liquidlytraded, so sometimes Reuters give us options prices that were traded say 1month ago. And our system was still using that price to get the implied vol.Because the ?price? of the option didn?t go down in the past month, we?re actually getting a higher than usual implied vol.If you meet such problems, how do you get an implied vol for the option?Using interpolation/extrapolation? Interpolation/Extrapolation may be a goodsolution for a single name, what if you have a big portfolio, I don?tthink this is a good solution. could you give me some advice?Thank you thank you

Everyone has this problem, but there is no simple fix.If you have available bid and ask prices, using the bid ask midpoint ( =(bid+ask)/2 ) might give a better IV estimate than using last price. Even so, some bid ask quotes may be stale.If you print out the number of contracts traded today, people will be able to see what contracts are illiquid (zero contracts traded) and can then ignore the IV information (you might even suppress the IV calculation in such cases and print 'N/A')I don't like the idea of an extrapolation scheme that makes up plausible IV data for illiquid options. When a portfolio includes illiquid options everyone from traders to managers must be aware of this, it cannot just be hidden by making up numbers IMHO.

QuoteOriginally posted by: acastaldoIf you have available bid and ask prices, using the bid ask midpoint ( =(bid+ask)/2 ) might give a better IV estimate than using last price. Even so, some bid ask quotes may be stale.It might be better to take the mid-point of the ask IV and the bid IV than to take the IV of the mid-point of the bid and ask prices.

It really depends what you'll use the volatility figures for. If it's to mark to market an existing position, then you have to assume some king of interpolation / extrapolation from the latest reliable market data. If it's to generate fair values that you use for trading it can be more dangerous as you get into the illiquid area(s) of the volatility surface you are considering. In that case it will come down mostly to the robustness of the volatility model you are using and its capability of fitting incomplete data.

A widespread solution is to calibrate a stochastic volatility model such as SABR. In this case your implied vols will be "fitted" by the model and not exactly Marked-to-Market. The advantage is that you can keep your model parameters constant or semi constant when there are no quoted prices. In the case of the SABR leaving your parameters unchanged and only adjusting your ATM vol level can take you quite far. Of course this opens a whole new universe of debate about which is the best model or technique for extrapolating OTM prices (or which quoted instruments will allow you to infer the distribution of the tails).