Suppose I have to price options on an instrument for which there is no existing options markets or surface (on multiple strikes, multiple maturities). I have only a lengthly time-series of OHLC prices (along with additional volume information). For example, since there is no existing options markets / surface, I can't use models like local volatility models. Have there been any studies/research on assets like S&P 500, Gold, Crude Oil or Bonds that have compared (i) actual impled volatility surfaces versus (ii) volatilities calculated purely using price data (e.g. Close to Close, EWMA, other non-linear alternatives) and tried to see which volatilities calculated purely based on price data come close (even if weakly close) to actual implied volatility surfaces ?

I would be wary of any such study. The problem is: the historical time series of returns *alone* don't tell you the systematic risks in the instrument. For that, you need to do some regressions/correlations vs major asset categories. Those will be important in trying to estimate the sign of the ATM skew.

QuoteOriginally posted by: AlanI would be wary of any such study. The problem is: the historical time series of returns *alone* don't tell you the systematic risks in the instrument. For that, you need to do some regressions/correlations vs major asset categories. Those will be important in trying to estimate the sign of the ATM skew.The instrument is likely to have 0.1 correlation to equity markets during bull markets and 0.3 during bear markets, but deviates in direction quite often. I'm actually looking for studies for exiting instruments like S&P that compared *instantaneous* volatility or ATM volatility with various measures of historic volatility to see what comes close.

QuoteOriginally posted by: pb273QuoteOriginally posted by: AlanI would be wary of any such study. The problem is: the historical time series of returns *alone* don't tell you the systematic risks in the instrument. For that, you need to do some regressions/correlations vs major asset categories. Those will be important in trying to estimate the sign of the ATM skew.The instrument is likely to have 0.1 correlation to equity markets during bull markets and 0.3 during bear markets, but deviates in direction quite often. I'm actually looking for studies for exiting instruments like S&P that compared *instantaneous* volatility or ATM volatility with various measures of historic volatility to see what comes close.OK. Not aware of any such. Some more unsolicited reaction.The trick is, I think, to get a predictor of ATM vol that varies in time as a true ATM vol would vary.Historical measures tend not to vary nearly as fast as reality. This suggests models which are largely driven by the ATM vols of related instruments with options: SPX, other securities in the same industry, etc.Also, you haven't said if it is a single-name stock. Any measure of historical volatility will behave in exactly the wrong-way through an earnings release.

when you have to price an option on an underlying with no traded options market, one thing to try to implement is a "risk-control" or "volatility target" style mechanism which dynamically allocates between your asset and some (assumed) riskless asset in order to control the realised volatility of this synthetic index. in doing so, most of the vega is removed from your option (i.e. with suitable overhedges, you no longer have to implement complex risk models to price options on it). obviously there are many cases where you cannot try this, but worth a shot otherwise.

The way I would approach this is through historical hedging simulation. After all, vol is supposed to represent the cost of delta-hedging an option. So I would start by running hedging simulation pricing the option for a given ATM vol to begin with and optimise that to find the one that gives me the best theta-gamma balance.

QuoteOriginally posted by: woodsdevilThe way I would approach this is through historical hedging simulation. After all, vol is supposed to represent the cost of delta-hedging an option. So I would start by running hedging simulation pricing the option for a given ATM vol to begin with and optimise that to find the one that gives me the best theta-gamma balance."if all you have is a hammer, everything looks like a nail"