Are there any papers that make an explicit contrast/comparison of the following (or other) vol models in terms of the suitability for addressing some empirical problem?

  • Wavelet multiresolution volatility
  • EWMA
  • GARCH family
  • range volatility
  • derivative implied volatility, models from stochastic finance. SABR etc.

You may want to first broadly categorize volatility models before comparing between them within each class, it does not make sense to compare standard deviation models with an implied vol model.

I would broadly classify as follows:

  • Historical realized volatility: Those include standard deviation (sum of squared deviations), realized range volatility models, and essentially anything that is based on past price and return data. Such models strictly deal with past data points and do not bother to make any sort of prediction.

  • Implied volatility models: Those lead to volatility measures that are the other side of the coin of derivative prices, whereas the implied vol model functions as translation tool. Part of that categorization is the SABR model and essentially most all stochastic models that implement different sorts of Brownian Motion "drivers".

  • Volatility forecasting models generally utilize past data, contrary to implied volatility models, in order to make predictions about future volatility dynamics and levels. Most Garch volatility models fall under this category.

  • A sub-category deals with intra-day volatility models such as Pearson or Garman-Klass.

I understand that my answer does not compare each individual volatility model but I believe, or hope, it helps to broadly compare and classify different models.

  • 4
    $\begingroup$ @Jase: adding to Matt Wolf excellent answer I would say that any model has at least one purpose, so comparing models should primarily take this into account. Best regards. $\endgroup$ – TheBridge May 23 '13 at 20:58
  • $\begingroup$ Very nice answer, but could you please elaborate on "it does not make sense to compare standard deviation models with an implied vol model." Why doesn't this make sense? $\endgroup$ – Jase May 24 '13 at 11:57
  • $\begingroup$ @Jase, because those are different tools aiming to target entirely different questions. (Sample) standard deviation measures variation in past price/return data whereas implied volatility expresses expectations of future asset/asset return variation. $\endgroup$ – Matt May 24 '13 at 15:30
  • $\begingroup$ What about comparing forecasts from a realized model to the implied vol? $\endgroup$ – Jase May 25 '13 at 1:37
  • 1
    $\begingroup$ @Jase, you can compare anything, whether you derive meaningful conclusions from your comparison is a different issue. I do not believe one derives value from comparing volatility models that are almost entirely based on historical prices with a model that captures market expectations of future price levels and/or variation. One is used to assess past price return variation, the other to price contracts with future contingent payoffs. $\endgroup$ – Matt May 25 '13 at 7:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.