This is not an easy post to write.
Not that I’m going to divulge some dark secret tucked away in the dank, deep corner of my alleged heart; it’s somewhat worse, because on this semi personal blog where I spew mind farts, I have to make the uncomfortable act of proving myself wrong – an act as pleasurable as substituting toilet paper for durian husks or using the ancient Japanese version of toilet paper. No wonder they came up with the ‘no paper’ toilet; ancestral memories sure last long.
What brought me here?
After reading NN Taleb’s Fooled By Randomness, (which I heartily recommend to any investor with a literary mind) I found myself faced with the growing realization that past performance is next to useless when it comes to predicting future performance.
Nearly every study done to determine if past performance can determine which fund will outperform finds little to no connection – the top performers of one year will be the bottom performers of the next year. No one outperforms all the time.
So far, this is known to me. I’ve always worked on the understanding that the best are right only 6 out of 10 times. The full quote comes from Peter Lynch, which reads,
"In this business, if you're good, you're right six times out of 10. You're never going to be right nine times out of 10." – Peter Lynch, One Up On Wall Street (?)
Where this becomes a problem is at the level of fund selection, and you have two funds, A and B, one of which has better 5-year annualized returns than the other. Typically, I would say that 5-year annualized outperformance means the fund has good long-term outperformance.
That’s what I’ve been told, that’s what I’ve been repeating. And, that’s where I’ve been doing it all wrong.
The problem is 5-years is still one single period – a long period nonetheless, but it’s just one, single period. I think the technical term for what this results in is small sample bias (or hasty generalization, or unrepresentative population, depending on how pedantic you want to be). Looking at various periods, say 1, 3 and 5 year returns, is equivalent to looking at 40 rich people in Singapore, and concluding that all people in Singapore are rich. Using three observations to arrive at a conclusion that applies to a fund’s general performance is likely to have more than a few blind spots.
So what now?
The answer, I think, will be to consider all possible periods, within a defined time. I’m referring to rolling periods, which accomplishes two things 1) increases sample size, which also 2) reduces end point bias.
And this brings me to the crux of the conundrum – a flawed recommendation.
Some time ago, I wrote about the DWS China Equity Fund, saying “The fund tends to outperform in up and down markets”. Ah youthful folly. Markets have since then pummeled the fund’s performance into mediocrity at best, and depressed ennui at worst.
Looking at its factsheet shows the fund returned (over 12 months as at 31 Jul 2012) -14.99% versus the benchmark MSCI China return of -10.96%, which is underperformance of -4.03pp.
Sure the China market has had a tough run, but actual performance is hardly consistent with my earlier statement. Clearly, in a toss-up between my conclusion and the market, market wins. The market is always right – even when it’s wrong. So the next thing to do is fix my conclusion.
Fixing Leaky Conclusions
Substituting rolling performances for discrete performances should at least help solve the issue of small samples. The fund’s performance will be compared against the benchmark. In this case, the benchmark is the Hang Seng. The Hang Seng isn’t entirely the best benchmark to use, but for this trial run, it’ll do.
Performance is based on all rolling 12-month performances from end-2005 to end-May 2012 (around 6 years). I want to separate market performance into 5 categories:
- Strong bull: greater than average positive market return
- Weak bull: within average positive market return
- Trendless: Zero
- Weak bear: within average negative market return
- Strong bear: greater than average negative market return
Over the period, the average positive market return is 22.6%, while the average negative market return is -18.47%. Across these categories, I’ll examine how the fund performs. The purpose of this is to help me arrive at a conclusion about how well the fund’s strategy/fund manager/team holds up across various market conditions.
That conclusion will be based on two key questions: 1) How often does the fund manager beat the benchmark’s return and 2) by how much does the fund manager beat the benchmark return?
Two charts answer the question:
Any conclusions based on past performance are made on the assumption that you’ll see roughly similar performances going forward.
Conclusion: Recent strong bear underperformance is definitely higher than average (-4.03pp, as stated way above), but not so much that it’s a huge difference from the average of -2.93pp which is good in the sense that performance numbers suggest the manager isn’t drifting too far from the established method.
Historically, this fund shines (and how) in a bull market for this fund to shine, but in a bear market, this fund will probably drift down to the middle of the pack, and possibly even further. Which it has.
For now at least, it feels like this is a better position to take. Sure it’s not the kind of certainty that people tend to believe strongly in, but I feel more comfortable with this position than I did in my previous one. For now.