GARCH vs Implied Volatility

Discussion in 'Options' started by TheBigShort, May 16, 2018.

  1. From experience which is the most efficient. I always find the GARCH family models over estimate realized vol. Heres a great website to use for vol models. Also when delta hedging, when we use the close to close model we delta hedge EOD, or EOW whatever your time period is. When we use the Garman Klass model we hedge intraday to capture the log(hi/lo) but what about the GARCH family? When do we hedge to realize GARCH vol? Screen Shot 2018-05-16 at 9.12.37 PM.png
  2. tommcginnis


    Many issues with these opening statements, but first maybe, is that since a common σ calculation is but a special case of a GARCH set-up (for example, the limit as measured heteroskedasticity tends to zero), and any GARCH-calculated σ is going to carry biases by construction. Why would you expect otherwise?

    Second, realized vol (whether from a common σ or as some GARCH byproduct) is an outcome computed from history. It is not an estimate; it is realized. If you wish to draw differences between the GARCH product and the common σ, have at it[!!] But unless you are specifying a population of data, and declaring a subset as the relevant lens, you have a population parameter, not a sample statistic.

    Third (and of related) concern: "When do we hedge to realize GARCH vol?" is not a statistically valid statement. It is the market which moves, and the market through that movement which defines volatility. We only realize mistakes. (If we're lucky.)

    Best wishes.
  3. But can't realized vol depend on the model and time frame you use? Ie. Crude could have a low close to close vol because it always closed close to the open but had a wild trading day. Therefor hedging intraday would have made better returns than hedging EOD using close to close.

    Also in the picture I have shown Vlab is using the EGARCH to calculate FB vol for next week which is 32%. Now EGARCH is supposed to be one of the most efficient vol models for forecasts under 1 week. FB IV for next week is 20%... What is going on here? Even if the EGARCH model was biased to its 60 day vol period of 38%, don't you think that IV to EGARCH spread is pretty high?

    Disclaimer: I bot FB premium today.
  4. tommcginnis


    Are you asking if I agree with the statements I just made to you? I do agree with myself.

    There is no basis for a relativistic conclusion like "pretty high," since we have no historic track of this 'spread', and we don't even know how this EGARCH was computed. That being said, if you compared IV and historic price movement of the S&P over the majority of the trading days back to mid-January, you'd find that IV<HV for most of that time.

    Slow down a little, spend a dollar, and buy a clue.
  5. upload_2018-5-16_22-44-28.png

    Just to confirm one statement. Lower study is 30 day intervals from YTD of Implied (VIX) and RealizedFutureVolatility.
    tommcginnis likes this.
  6. would you mind sharing the tos code?
  7. your great, just gave you a follow on tos. Thanks for providing. I will tell you if I find any problems with it
  8. is there an error in that? I cant seem to beable to download it. Downloaded one of your other ones no probelm
  9. #10     May 17, 2018