Hey , B... totally agree with you , stat vols are irrelevant , especially without assigning weights to reporting/non reporting month per basket's component.
Is partial replication a reasonable risk ? If so, this then begs the question of which stocks to select for a partial replication ? How is the optimal basket defined ? Obviously heavily weighted stocks, but what percentage of the index would you reasonably need to replicate ? What about sector representation ? What about differing volatilities of 2 stocks that operate in the same sector, should that be a consideration to the basket weighting ? What values of implied index correlation would you run a dispersion trade ? Obviously +1 is a no-brainer, but what about +0.90 or +0.80, where is the cut-off ? Same question for a reverse dispersion where implied index correlation is 0.00 being a good risk, but what about +0.1 or +0.2, where is the cut-off for a reverse dispersion ? Correlation â do you guys use the Pearson methodology or the Spearman rank calculation ? Which is more appropriate for measuring stock correlation ? When you get the time.....thanks again.
Again you don't tell us what dispersion strategy you are referring to so I'll have to guess and say it looks like -dispersion where you are selling the components, but I really can't be sure since there are different philosophies about using correlation to forecast. The IV/SV ratio of the 2 stocks is the same, so I don't see how you could compare them. I also don't know where you get MO correlation as 16.5%? Ivolatility is showing it as 3.92%. I believe you need to either roll your own stats or know how your service is calculating them and then check it to make sure. I think that correlation is the fly-in-the-ointment of dispersion trading. More than volatility and even more than price, correlation seems to jump all over the place. And there are competing thoughts on how to calculate correlation. I have attached a chart of 30 day correlation of AIG and MO over the past 30 days. This is the Pearson correlation of log returns. If it was the Spearman rank correlation you might really see some movement.
There is a lot of factors to consider when making decision to exclude the stock for partial replication like Beta , IV/HV ratio , correl , deferent stock's-to-Index weights, reporting [none] and more... Many factors are contradicting each other , so I want to use as few as possible. For "traditional" (sell Index) dispersion , I'm leaning toward High Correl + High IV + Reporting/No as heavy factors to Exclude , while still maintaining atleast 70% of basket to Index weight ( about 20 stocks for DOW). Mistic , I was referring to "buy the components" scenario in my post. Why are the MO's correl is 4% ? Your file shows it around 17% (same as Egar and my own calcs).
IV_Trader, when you say that Egar shows MO correlation at 17%, what exactly are you referring to as your source for that? Are you an EGAR subscriber? You have also been doing full replication on the Dow, so why the change? Because of its price weighting, the Dow is not particularly suited to partial replicaiton, although other factors such as beta may come into play. Yes, there are a lot of factors to consider in dispersion. The first fractor is the time frame of your trade. If it is to be greater than 1 month, then probably volatility would be more important. Thus for regular dispersion the IV should not be high and the IV/HV ratio should be looked at. This would be similar to the criteria for a good straddle. On the other hand, if your time frame is 1 month or less, then correlation becomes more important than volatility (gamma more important than vega). Then what EGAR calls the "third volatility level coefficient" becomes most important. The main formula is then calculated with price correlations and HV rather than IV and component IV correlations. Beta is also important as the combination of volatilty and correlation, but it would need to be calculated yourself short term. 5 year betas given out by almost all sources are useless.
The dow is an easy replication pair due to price weighting. I think you guys are suffering from "quantitis"
yes , I took 14 days Egar trial (and NOT impressed at all , but it's off subject now). Correct , I only did the full replication so far , but after reading posts on this thread , I am looking into partial replication (currently testing). Thanks for your post , I will look at Beta too.
Current Basket-IV / Index IV ratio=2 , I will go with reverse dispersion for JAN again : Long DOW stocks Sell slightly ITM components calls (ratio is 1:2) Buy DIA ATM puts , where premium received from calls = premium paid for puts. Worked well for me last month.
You realize you'll be short the [slightly] bull delta synthetic straddle with your components? Short the +delta straddle/long index puts. You may want to consider fronting the cheap-curvature in buying the deep calls. You'd have a replicated iron fly.