Wednesday, 29 August 2012

Surviving Survivorship Bias

I spent about 5 days in the sleepy town of Kota Kinabalu, two of which were spent climbing Mount Kinabalu. I wish I could say I sprung up the mountain and posed like the Baboon in the Lion King as wildlife grovelled at my feet. Closer to reality it was a drunken stumble with pink elephants on parade.

On a sidenote, I think “pink elephants on parade” is one piece of supporting evidence for legalizing drugs for artistic/commercial purposes. Just an opinion.

But apart from climbing mountains in a semi-delusional state, I found the time to finish a first read of Nassim Nicholas Taleb’s “Fooled By Randomness” (2nd ed). The essential argument I took away from the first read through, was that markets are literally more random than you can think. I’d use the word imagine, but imagination is often the aborted fetus of logic-driven investment decisions, so the burden of comprehending the randomness of markets will sit squarely on the shoulders of conscious thought.

One discussion that struck me was two chapters devoted to illustrating and discussing the concept of survivorship bias.

Survivorship bias, as defined by Investopedia, is:
The tendency for mutual funds with poor performance to be dropped by mutual fund companies, generally because of poor results or low asset accumulation. This phenomenon, which is widespread in the fund industry, results in an overestimation of the past returns of mutual funds.
This is relevant to my portfolio, since I do tend to look closely at the performance of funds when I go about selecting them. So what’s to be done about this?

Some data providers have created survivorship bias free indices, which includes constituents (like funds, stocks, bonds, derivatives…pretty much any asset that can be invested in, will be) that have been merged/acquired/closed. It’s a lot of effort to track dead assets and one would have to ask, does it matter and if so, how much?

 In the case of a stock market, it’s known that index constituents remain constituents only if they meet a set of criteria, which includes track record, profitability (pre-tax losses for 3 recent consecutive financial years), and minimum market capitalization (an average daily marketcap of S$40m over the last 120 days). Link to SGX rule book here. This is the essence of survivorship bias – any company that falls below the criteria is automatically filtered out, leaving only ‘survivors’. No wonder stock markets have an upward bias.

In the case of funds, if they fall below a certain amount of assets, then there’s a good chance the fund will merge. If a fund manager consistently underperforms the market, then there’s a good chance the assets will be transferred to another manager’s watch, leaving only ‘survivors’. So mutual funds have an upward bias too – good managers stay, bad managers drop out.

So as long as one sticks around long enough without getting booted off the game, you’d considered a winner, even if it means hugging the benchmark like a python with separation issues. From a past performance perspective, this suggests that looking at past performance to determine a manager’s returns are due to luck (going along for wherever the market feels like taking me) or skill (sorta like luck, but requires more preparation) is of limited value. Which is unfortunate, because past performance is often all one has to go on. According to this article most academic literature has only demonstrated fund manager returns are more likely due to luck rather than skill. As far as I know, no one has identified how to separate the Homers from the homers.

So if fund selection based on performance is of little to no value, where does that leave me? Down to the elements I can control – asset allocation, selection and diversification. I would love to find a method that helps me identify fund managers with a consistent strategy that will over time beat markets, but I’m not sure if it exists. And even if it did, in all honesty, if it comes from an academic journal, I’m not sure I would find it a) understandable or b) applicable…especially if it’s arcane academia written by Taleb.

He looks better bald and rich