wilmott.com

Does higher implied correlation in the indices generally "challenge" price-mean reverting equity pairs books? It seems that higher correlation would "prolong" the amount of time that stocks need to mean revert...

I'm not sure I understand you, but I think the answer is no. A high implied correlation of an index (that is, options on the index are more expensive than you would expect given the prices of options on its components and measured correlations among the components) suggests that pairwise correlations are higher than historical values, not lower.More important, pair trades are based on complex local phenomena; index implied correlation is a simpler, global thing. It's hard to reason much from one to the other, like if I told you the mean global temperature tomorrow would be 0.1 degree C hotter than expected and asked you if that meant it was more likely to rain in Brooklyn.

I suggested higher implied correlations to infer that correlations are up across the board -- and I now realize this -- Let me start over by first noting that in my original post I made a typo: 'would "prolong"' should have been 'wouldn't "prolong"'.Redux:Pairs trades seem quite paraochial and subject to the "local" events inherent in each stock within a pair. Pairs trades entered into at some "divergent" trigger point should do better if it's component's log returns series move more in-line with each other. Now let's assume the market suffers a shock of some sort that really spikes correlations across the board: can it be assumed that this would "supercharge" the return in a pairs book? Would a large diverse pairs book covering many industries always benefit from a true pop in implied correlation? My hunch is that the aforementioned ideosyncratic nature kicks in to mute any such exceptional performance... Am I right on this?

My hunch is that a spike in market correlation means that there's a single market-wide factor that has become temporarily important. For example, last week's employment number was closely watched for a variety of reasons. That induced a short term high correlation among many stocks around the time the number came out. I don't that that affects pairs trading one way or the other. It's true that the pairwise correlation between the pair goes up temporarily, but only because of increased volatility. What you really care about is the residual volatility:where C is the covariance. A common factor added to both x and y, independent of x and y, will not affect this. The increase in the variances will be exactly offset by the increase in the covariance.Another important point is that pairs trading should not be based on correlation. Correlation measures only contemporaneous price comovments (the only kind allowed in efficient markets). You can't trade on that. Pairs trading works because the comovements are not always exactly contemporaneous. If stock A goes up $1 we might expect that stock B will go up $0.50 in the next week. If we buy B and short A we capture that expected profit and hedge away most of the residual risk. The prices of stocks A and B will cointegrate more than their correlation suggests.

... which leads us to vector error correction model. So am I correct in saying that, assuming said 'exogenous macro event of noteworthy repute' isn't a long-term, redefining regime change for the economy, the marketplace or more specifically for stock A or stock B, the true characteristics of that pair's cointegration will most likely remain unfazed by the shock. Thus there is no "elongation" of the expected time-to-mean reversion as posed via the VECM. Right?

I think I agree with you.

Thanks, Aaron!