Earnings IV Decomposition w/ index correlation

Discussion in 'Options' started by tooriginal, Oct 4, 2022.

  1. upload_2022-10-4_22-36-52.png


    upload_2022-10-4_22-37-39.png

    upload_2022-10-4_22-38-16.png
    10 years of daily returns. Also made sure to do 30 days of 5 min returns etc.
    upload_2022-10-4_22-42-53.png

    Daily (as opposed to annualized) IV
    upload_2022-10-4_22-43-17.png


    Volatility is not additive so (implied) variance is displayed instead. Holes where the non earnings regression related variance exceeds the implied variance??
    upload_2022-10-4_22-44-20.png


    Calculated using the fact that earnings is two trading days before expiry:
    upload_2022-10-4_22-46-16.png

    My interpretation is that if the model holds any water, it is in disagreement with the market about the quality of the regression going forward as the beta variance + residual variance can be larger than the implied variance at which point the earnings variance would have to be negative.

    If the model is worthwhile the straddles could be a buy here.

    A dumb model or what?
     
  2. It's dumb,
    Trading is unlike creating, inventing, discovering a scientific, linear, fixed, nuclear bomb formula....the activity of the market is rather part art, part science,
    To trade can be a rather poetic, dynamic, process....you have to embrace science/data, patience, calmness, combined with current movements, ranges, and expectations, people, emotions, thoughts, and where things should, most likely, inevitably be and end up at,
     
    Last edited: Oct 4, 2022
  3. What's the goal of this model or this kind of research?
    To predict the change in IV (or HV) when the quarterly Earnings Report of the company gets released?
     
  4. Matt_ORATS

    Matt_ORATS Sponsor

    Here is how ORATS does it.
    We determine the IV of each month through our strike smoothing process. (atmiv)
    We perform a term structure fitting function with three variables, short term, long term, and earnings effect, and continue until we minimize the squared error. (calVol)
    By removing the earnings effect we can observe the non earnings IV, (unadjVol) and the amount of IV that is related to earnings (earnEffect).
    [​IMG]
    https://gyazo.com/4e274c7e4804432622b2059125edd721

    The above uses an index but that can be misleading.
     
    earth_imperator likes this.

  5. It looks like you are taking differences even though sqrts don't commute with subtraction? Consequently earnEffect will be increasing as time passes and earnings become a proportionally larger component in the average so it doesn't directly help you know how much of the earnings contracts (10/21 expiry) are actually attributed to earnings. Correct?

    Moreover if stock Y is correlated to index X, and if the market views index X as having more volatility in the near term then the far, then it follows that, depending on the strength of the correlation, the market would be inconsistent to not view stock Y as having more volatility in the near term then the far. Propagating the correlation through the term structure can be troubling due to the volatility of the beta and R^2 however. Do you agree?
     
    Matt_ORATS likes this.

  6. The least squares against all expiries makes more sense then a closed form algebraic solution. Thank you.
     
  7. Matt_ORATS

    Matt_ORATS Sponsor

    It makes more sense to solve for earnings effect and result in a rational term structure than to result in a term structure that matches an index.