Sunday, October 9, 2011

Survivorship Bias, Hedge Funds and Bad Research

On Friday I challenged myself to read this "research" by Pertrac that claims to identify the impact of fund size and age on hedge fund performance.  In that post, I gave two very simple explanations for why small and young hedge funds might perform better.

Explanation one: People start hedge funds with a really good idea.  When that above average idea has played out, they become average.

Explanation two: Small funds, executing the same strategy as large funds, can operate in relatively less liquid parts of a market, therefore appearing to perform better when they're really collecting a liquidity premium.

More importantly, however, survivorship bias will explain that result too.  And even more results.

I wrote the post without reading the paper.  Now I've read the paper.  It's worse than I expected!

The authors actually imposed more bias on the data in an attempt to clean it!  Page 10 says they removed 311 funds that did not report December 2010 results.  Who fails to report numbers?  I doubt top performers fail to report voluntarily.  In the section "III. A Final Check on The Dead" I figured I was dead meat...they somehow corrected for survivorship. How wrong I was!  This section details how they carefully removed all the data about funds they believed had died.

So, now we know that the study carefully analyzes a large database of funds that has just as "carefully" removed from study the funds that have died either during the year in question or during prior years.  I would have no problem with this removal of data if we had any reason to believe the deaths are random!  However, hedge fund death is nearly entirely performance driven (if not actual, than at least fraudulent performance in the whole portfolio was invested in a fraud.)

Next post I will discuss the only four potentially interesting statements in the Pertrac study.  Here's a hint: They all involve statements about volatility.