Wednesday, October 11, 2006

In Search of Dark Alpha

Much of the current focus of my ETF analysis has been been around an idea I've been referring to as "Dark Alpha." Alpha, according to the classic definition, is the extra return a security provides that is not explained by its correlation to beta (as defined by the correlation to the reference index; normally the S&P 500 but occasionally EEM is a little more of value in emerging market analysis).

So if the S&P goes up 1% and EWJ (iShares Japan) goes up .99%, S&P goes down .5% and EWJ goes down .498% and this tracking stays very constant, we have a pretty good definition of a positive beta close to 1.00 (as well as a good value for R-squared, which tells us statistically how strong that beta relationship is between the two per explaining changes in values).

Alpha seekers look for gains that occur on a particular security that are beyond that which the beta correlation predicts. If you think of the standard deviation fluctuation of the stock (the up and down movement over time as plotted on a graph), alpha seekers are looking for that extra unexplained boost on the peaks. It's like mountain climbing in a sense, looking for summits that are taller than the laws of physics might predict.

On the flip side of mountain climbing is spelunking - typically a place where people long in the market don't really enjoy being. It's the bottom part of the graph, below the opening line. On our stock chart, beta predicts a correlation that is the full standard deviation round trip. If the S&P 500 has a rather bad day and goes down a full standard deviation, your beta correlation should stay true. You should experience the same correlation to beta when things go down as they do when they go up.

However, just as the alpha summit seekers find "extra tall peaks" in some stocks, there are oddities on the other side of the chart, down in the caverns where bears reside. Akin to the concept of dark matter, deep in the caverns resides Dark Alpha. It's the unexplained non-loss that is not explained by beta. Technically, a non-loss is still a return, and one could argue that dark alpha really is alpha in that respect. However, there is a reason for splitting dark alpha apart from its cousin.

When stock markets go significantly under the line (say, at least 1 standard deviation), the go into stress. This is interesting to some in that markets have been observed to misbehave under stress. Specifically, there's the saying that "under stress, correlations approach 1.0." That's a clever way of saying your highly diversified, low correlating portfolio suddenly decides to act like a single stock on days where the market really does poorly.

But emerging market ETF analysis on "correlation under stress" shows this isn't always the case. Using a more aggressive stress model (using normalized z-scores < -2.0 for the index and then running a regression on various ETFs to the stressed index), there are some interesting results. Some markets immediately approach that 1.0, confirming our hunch. But others don't, and stay relatively unstressed.

What that means to investors is that some emerging market indexes provide much greater diversification, as well as potential resistance to high stress global events, which is of significant value. As many add emerging market exposure to their portfolio for this very reason, it's useful to test if this insurance policy is really worth anything at all, as well as identify which policies do a better job than others.

Of course, that data only points the direction of where to look further. My initial hunch is that we really have to think of what is in the ETF index once again (as my energy sensitivity analysis has shown - e.g. own an ETF like iShares MSCI Brazil [EWZ] that has more than 60% direct exposure to energy markets in its composition and you'll not surprisingly have an equity that behaves like an energy equity, not a "Brazil" equity whatever that would be).

My next steps include re-running the models under less stress ( Index daily Z-scores < -1.0) to see the strength of the correlation there, and then move into ETF decomposition to see if we can identify some sectors within that are showing the immunity to beta-induced stress. Already, global financial sector holdings show some interesting resistance. If that holds true, it may be some of those bank vaults in Singapore and Austria are holding secret deposits of dark alpha.


No comments:

Recent Posts

Older Posts