Naturally, the main attraction of market neutral funds is that their performance is not correlated with markets.  But usually the term “market neutral” refers to the current positions held in the fund, not its actual performance vs. markets.  The aggregate beta-weighted exposure of a hedge fund is rarely the same as the eventual market beta of the fund’s track record.  In fact, these numbers are usually quite different.  Paying too much attention to the holdings in the portfolio at the expense of the fund’s overall performance is like missing the proverbial forest for the trees.

Market neutral hedge fund managers will often talk about a certain amount of beta that will “work its way into” their funds.  They say it’s inevitable given the large number of trades they makes and the plethora of factors they try to neutralize.

But how exactly does this occur?  After all, if a long/short equity fund were perfectly market neutral (e.g. long $100 million and short $100 million), then why would its return stream suggest it is actually long (or short) bias when compared to the market?  Managers are quick to point out that they may have a beta-weighted bias.  In other words, their shorts might have a higher beta than their longs – they may be ”short bias on a beta-weighted basis” (as opposed to a dollar-weighted basis).

That’s fair enough.  But what if a beta-neutral fund’s return stream still shows a significant non-zero market correlation?  “Well”, the manager might say, “That’s because we’re not sector neutral.  Yes, our dollars and betas add up to zero.  But we still may be net long consumer non-cyclicals and net short financials.”

Being the quants that they are, market neutral managers often obsess with perfect neutrality – a Zen-like state where the world is in perfect balance.  So let’s say they neutralize the above sector biases and more.  Then why on Earth would a fund that is dollar, beta, sector, interest rate, and market cap neutral have anything other than a 0.00 beta?

The answer lies in the r-squared’s of the betas being used to calculate static “beta-weighted neutrality”.  Let’s take a simple example of a fund with two positions: $1 long Stock “A” (1.0 beta to equities) and $1 short Stock “B” (1.0 beta to equities).

As we’ve discussed on this blog, beta is the product of two numbers: the correlation (r) and the volatility relative to the market’s volatility (i.e. volatility as a percentage of the market’s volatility).  Let’s further assume that the beta of Stock “A” has an r-squared of 0.25 and the beta of Stock “B” also has an r-squared of 0.25.  Basically, we’re not sure exactly how either stock will behave vs. the market as their correlations (r) are both only 0.5.

With correlations like these, it’s no wonder that the actual performance of “A” and “B” does not live up to the “1.0 beta” billing!  The actual performance of “A” and “B” will be normally distributed around the value suggested by the 1.0 beta (i.e. the market return).  So in the long run, with hundreds of data points, the actual returns will approach what the position-level beta had predicted (assuming a commensurately long look-back window was used to calculate the betas).  But in the short run, this likely won’t be the case.  A manager’s ability to predict how their holdings will behave in the short run is what will determine how closely their track record’s beta will match his position-level beta-weighted net exposure.

For example, if a manager was consistently net long 20% on a beta-weighted basis for 1 year, but managed to post returns that had a zero market correlation, this would suggest she had correctly guessed that the position betas she was viewing each month were actually wrong.  Her Bloomberg was telling her she was net long on a beta-weighted basis.  But perhaps she knew better.  She knew that the betas going forward were actually going to be lower than what had been observed in the 60 or 90 day look-back window commonly calculated by tools like Bloomberg.  In other words, maybe she knew that she was actually market neutral all along.

Or maybe she was lucky.  Who knows?  Our point here is only that the difference between static beta-weighted net exposure and the eventual beta of the fund itself can be explained by the r-squared’s (correlations) of the betas being used to calculate said (static) market neutrality.

As an aside, the only way a fund’s position-level beta-weighted net exposure would exactly match that fund’s eventual beta is if the r-squared’s for each holdings’ beta was 1.0 (i.e. correlations=1).  In other words, if the betas were all perfectly predictive.  This is obviously impossible.

So next time you talk to a “market neutral” manager, it might be worth asking if the fund’s aggregate positions are beta neutral or if the fund’s actual performance is beta neutral.  In other words, don’t miss the forest for the trees.

- Alpha Male