The hunt for the true investment guru
This is a revision of a post originally published November 21, 2019. See >>>Revision below.
Most investors are aware that the money manager or investor who has beaten their benchmarks the last year or two is considered by many to be a whiz. In looking at this question we have two thorny problems as humans. We see patterns in random data. We jump to conclusions from inadequate data.
The hunt for the investment guru is the least of our problems. The title was just to get your attention. What follows is generally applicable to every decision investors make based on data. And of course, most of our decisions are based on data; think of the performance of management, a stretch of earnings, company sales or any other metric in which human endeavor and luck are both involved.
Kahneman offers this test to his readers. Look at the following three sequences where B is the births of boys and G is the births of girls born in sequence in a hospital:
BBBGGG
GGGGGG
BGBBGB
He asks the question whether the sequences are equally likely. Look at them closely. One stands out. So, what is the answer? Are the sequences equally likely? Kahneman writes: “The intuitive answer – “of course not!” – is false. Because the events are independent and because the outcomes B and G are (approximately) equally likely, then any possible sequence of six births is a likely as any other. Even now that you know this conclusions is true, it remains counterintuitive, because only the third sequence appears random.” (Kahneman, 2011)p.115.
Now change the example slightly. Look at the same three sequences. But instead we track the investment performance of three investors. B is bad performance in the stock market – underperforming the average investor. G is good performance – outperforming the average investor performance. The first investor has a track record over six years of BBBGGG, that is, they had three bad years followed by three good years. Since our benchmark is the average performance of all investors, in any given year half will underperform and half will outperform. Our second investor had six good years!
Now the same question. Are all the sequences equally likely? Instinctively we want to say the second investor is the guru. More thought is required than the birth of boys and girls. There we agree the outcomes are approximately equally likely. As well, the birth of a boy vs the birth of a girl is completely random. Because of this, the sequences are equally likely.
It might be said that given the universe of investors, it is equally likely that any investor would be above or below average in any given year, just like the chances of a girl or boy. That would be the situation where any outperformance by an investor was simply random luck.
To assess whether our GGGGGG investor is a guru we need the help of statistics. We do not ask what the chances are of outperforming six years in a row. We also need to ignore our love of patterns. To use statistics we need enough annual data so that we can say the outperformance is statistically significant. Statistics talks in terms of being 95% sure the results are not due to pure luck, i.e. is due to skill. There are millions of investors. If we asked each one to flip a coin twenty times, we can suppose that some would get 20 heads in a row. Undoubtedly amongst the millions of investors there are a large number who have outperformed the market for twenty years purely by luck. The question is how many years of cumulative outperformance is necessary to be satisfied the performance is the result of skill. Someone has apparently figured out that a fund manager would need to beat the stock market for 36 years before we would know that his performance was based on skill rather than luck! I can’t vouch for the conclusion but I could believe it’s true.
In our hunt for the investment guru we get hit with a double whammy of cognitive errors. We see patterns in random data because we, as humans, love to find causes and patterns suggest causes. And, we jump to conclusions on the basis of statistically insignificant data, again, because we love to identify causes. So, can we learn much from a few observations? Not really. But there is a high risk that we will try. Forewarned is forearmed.
>>>Revision: The cognitive errors discussed in this post only touch the surface. Being mislead by pseudo statistics and even being mislead by sophisticated statistics comes up again and again for investors.
For example, what are we to make of the assertion by some pundit that the stock market has advanced 20% or more 17 times and that 14 of those occasions were followed by an advance the following year. Or how about the assertion that the most successful stock market investments, on average, are in small companies. These are examples of what Daniel Kahneman calls the Law of Small Numbers.
Actually, there are more statistical blunders than you can shake a stick at.
If 100 companies are looked at over a 20-year period and ten of the companies have gone bankrupt, it is wrong to use the remaining 90 alone to measure performance. This is the problem of survivorship bias in data sets.
Or take the case when a data series is thought to follow a normal distribution but in reality, follows a different distribution. For example, stock prices apparently tend to have longer tails than the normal distribution. This is the kind of problem that bedevils quants and was directly related to the demise of Long-Term Capital Management (LTCM). The collapse of LTCM was a spectacular failure of a highly leveraged hedge fund in 1998 that necessitated a $3.625 billion bailout orchestrated by the U.S. Federal Reserve to prevent a wider systemic financial crisis.
Here’s another. There is the common problem around back testing to prove various quant strategies. A thousand different strategies can be tested. Difficulties arise when cherry picking of results occurs. Some of the thousand tested strategies are bound to show up well in a back test. That doesn’t mean they will work well in future.
These back tests also suffer from the weakness of non-stationarity of data when the world is changing and what’s happened in the past is no guide to the future. For example, the famous CAPE suffers from this problem. Reported earnings in companies with major investment in intangibles of lasting value don’t mean the same as they used to.
And on the subject of patterns, random data, such as the results of flipping a coin one hundred times, are susceptible to recording and analyzing and graphing. The amazing thing about random numbers is that they can show apparent patterns. Of course, the patterns are only illusions. They aren’t telling us anything. It seems humans constantly fall into the trap of reading something into patterns in random numbers.
Daniel Kahneman explains: “Random processes produce many sequences that convince people that the process is not random after all.” (Kahneman, Thinking, Fast and Slow. 2011) p115 This is a risk we must be conscious of.
I discuss these problems in later posts titled The Law of Small Numbers for investors per Daniel Kahneman, CAPE’s halo falls with a thud and Random Walk fallacy in investing End Revison<<<
Other posts on investment psychology
This post is part of a series. Readers are invited to read Investment psychology explainer for Mr. Market – introduction This will give you a better understanding of some of the terms and ideas and give you links to other posts in the series.
Want to dig deeper into the principles behind successful investing?
Click here for the Motherlode – introduction.
If you like this blog, tell your friends about it
And don’t hesitate to provide comments
2 thoughts on “Let’s face it, we’re really bad at patterns and statistics”