Skip to content
All library documents

Trading Volatility Correlation Breakdowns with Options

Article Quant Q&A · Author: strimp099
Read the original

Summary

The discussion examines why 90-day realized volatility estimates for JPM and the S&P 500 diverged during August 2011, and whether that divergence could inform options trades or volatility forecasts. It attributes the difference partly to the assets’ prior price paths and partly to how the Garman–Klass estimator weights intraday ranges against close-to-close moves. Because JPM and the index had different return patterns, the same market stress did not affect their rolling estimates equally.

For listed options, the proposed expression is to buy gamma in one underlying and sell gamma in the other, provided a trader can forecast which asset’s volatility will outperform. The discussion says a correlation signal alone is insufficient: its persistence and the direction of the relative volatility move need testing. It also distinguishes realized from implied volatility, suggesting tests for autocorrelation and realized-implied dispersion when assessing possible forecasts. Earnings and bank-specific macroeconomic exposure are offered as other causes of correlation breaks. The example is an explanation, not evidence of a reliable trading edge; its claims require empirical validation.

Key ideas

  • Garman–Klass volatility responds to intraday ranges and close-to-close moves in different ways.
  • Different price histories can cause related assets’ rolling realized volatility estimates to diverge.
  • A correlation breakdown alone does not identify which asset’s volatility will outperform.
  • A listed-options expression may pair long gamma in one asset with short gamma in another.
  • Forecasting future volatility requires testing persistence and links between realized and implied measures.

Tags

EquitiesOptionsVolatilityStatistics
Full text
# Exploiting breakdowns in correlation of estimated volatility


# Exploiting breakdowns in correlation of estimated volatility












In the attached image I have a plot of the rolling correlation of 90-day historic volatility (using the Garman Klass estimator based on Sinclair's Volatility Trading) of JPM v. the S&P. As can be clearly seen, the correlation is generally somewhere near 1. However, there are times when the correlation degrades significantly (witness summer 2011). I'm curious of two things:

- Trading strategies that may be implemented to exploit the breakdown using listed equity options (assuming one expects the correlation to revert back near the historic norm, in this case near 1)

- How one might use the information to help predict future volatility

## Answer by Matt Wolf (score 3)

https://quant.stackexchange.com/a/3837

There is a simple explanation in this specific case (SPX vs JPM rolling 90-day realized vol). When you look at the specific daily price series of SPX and JPM during August 2011 (a very bad month for most hedge funds) you will notice 2 main differences between each series:

1) JPM was already in a down trend and thus negative returns (or lower closing prices vs. previous day closing prices) in August did not push up its 90-day rolling vol up by too much. On the other side traded at very lofty levels all the way into July 22 2011 when market participants started to dump shares. Thus the negative returns (lower prices) pushed up volatility in SPX by much more than the vol measure of JPM.

2) During the days when the market generated the biggest losses for both JPM, and SPX the key difference here is that JPM traded very large intraday ranges while the closed to previous close was only elevated at 3-4 occasions in August while SPX traded relatively similar large intraday ranges but the close to previous close was also very large. This leads me to an observation of the GK formula:

GK=0.5(ln(h/l)^2)-.39(ln(ct/ct-1)^2

where h=hi, l=lo, ct=close today, ct-1=close yest

You notice that GK favors high intraday ranges and penalizes for large close to close ranges. When factoring points 1) and 2) into the formula one sees that the effects are partially offsetting each other, but my hunch is that the change in vol profile in SPX (point 1) are overshadowing the higher GK vols in JPM due to relatively higher intraday vol vs close to close vols.

To get to your questions:

1) Someone pointed out you need to make sure you understand the differences between your historic vol measures and implied vols. You mentioned you look for strategies on the listed options side making this a lot more difficult to exploit. Let's say you are good at forecasting a breakdown in your specific correlation figures. This in itself does not give you anything. You need to test for auto correlation in order to benefit from such ability looking forward. Also evenif you correctly predict a breakdown in correlations you need to be right on which asset's vol profile outperforms the other (the sign of GK(SPX , 90) - GK (JPM, 90) for example). Given you are able to do so your only way through listed options is to trade gamma differentials. You buy gamma in one underlying and dell the gamma of the other. You are much luckier on the OTC side because you can trade correlation swaps, soft exotic correlation/dispersion baskets to express views in changes in correlation profiles.

2) I touched on it but your task is basically to test for autocorrelation. Observing historical variables has zero bearing on the future in the absence of auto correlation. Fortunately certain implied volatility metrics display high auto correlation but you need to test for it in a rigorous fashion. This is how many professional options traders utilize realized vols to make inferences on expected changes in implied vols. Some traders explicitly look at dispersion between realized and implied vols. But the key again here is that you need to establish a link between the two through auto correlation.

Hope this helps a bit.

## Answer by onlyvix.blogspot.com (score 1)

https://quant.stackexchange.com/a/3533

There may be many reasons for such correlation breakdowns - quarterly earnings announcements that result in elevated volatility for JPM but not SPX, or systematic risk: JPM being a bank has more Euro-debt related risk than SPX.

Having said that to trade options you should be looking at implied volatility time series.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.