A Better Way To Measure Performance Persistence

by: Christopher Holt

“Performance persistence” is never far from the minds of both hedge fund investors and researchers. Many studies have attempted to determine if good managers can actually remain good or if their performance is destined to “revert to the mean”. In general, they conclude that persistence does exist ... bad managers remain bad. Unfortunately, the opposite is not generally believed to be very true.

One author of a recent study on return persistence, Daniel Capocci, found that previous returns showed modest predictive ability and wrote on these pages last March:

“…we found that the alphas were significantly positive for deciles 2 to 8 in the previous year’s performance rankings, but not for the previous year’s best and worst performing funds.”

Now Markus Schmid and Samuel Manser of the Swiss Institute of Banking and Finance and Credit Suisse say they have identified what they say is a better way to measure persistence - by looking at only one strategy, long/short equity. Taking aim at Capocci and others, they argue that lopping hedge funds with different investment strategies into the same analysis is less robust. Analysing just one strategy, on the other hand, removes strategy-specific betas from the equation.

Like Capocci, they divide hedge funds into 10 portfolios depending on their previous performance. Then they track the performance of those portfolios going forward. Here’s what they found:

As you can see from this chart, the best performing funds during the “initial” period actually have a slightly higher chance of being in that category during the subsequent period. The worst performing funds during the initial period had at least as high a chance of remaining the worst performing - and triple the chance of dropping out of the database altogether.

Unfortunately, when you look closely at this chart you also see that the best funds in the initial period also have a relatively high change of totally bombing in the subsequent period (either ending up in the bottom decile or falling out of the database entirely). There’s also a solid chance that the real stinkers from an initial period can actually vault to the first decile in the subsequent period.

But despite this, there does appear to be some very modest evidence of persistence across different deciles. The chances of a 9th decile fund having been in that same decile during the previous period are above average (or more accurately, above the median). Same for funds in the 2nd, 3rd, 4th and 7th deciles.

Still, these relationships are weak, say the authors, and “forming portfolios based on lagged returns does not seem to be a useful way to detect alpha portfolios”. Plus, rapid changes in return ranking could simply be the result of being long- (or short-) bias during a big market swing.

So instead, they conduct essentially the same analysis using alpha instead of raw returns. The results are more promising:

Now we’re talking! You don’t need a microscope to see that relationship. The return persistence across the middle of this chart can easily be seen with the naked eye. Say the authors:

“…significant alphas are clearly located in the portfolios containing the funds with high lagged alpha. Persistence in alpha is especially pronounced for the extreme portfolios 1A and 10C. [ed: the best and worst initial-period portfolios].”

The authors conclude by echoing the caution that Capocci raised in his AAA guest posting: that return-chasing (or at least alpha-chasing) succumbs to the friction costs of lock-ups and trading costs.

But apparently, that doesn’t negate the significant value that can be added by a bit of pre-investment alpha analysis.