I read a book once called “What the Most Successful People Do Before Breakfast.” I read it unironically, which is why I am ashamed to admit it.
There was an article in Business Insider explaining why Amazon was so successful. The secret sauce is Jeff Bezos’ preference for speed of execution. The evidence is the wording of his first-ever job ad, which BI proudly published with these words highlighted: “you should be able to [build large and complex systems] in about one-third the time that most competent people think possible.”
This is a stupid argument.
To test the hypothesis (speed → success, measured by sales, profit, etc.) properly, you’d need to:
take a sample of startups that began around the same time as Amazon;
identify a bunch of independent variables (features), including the number of times speed of execution was mentioned in job ads;
run a multivariate regression (with sales, profit, etc. as the dependent variable) to see whether the speed variable is significant and carries the expected (positive) sign.
BI never bothered. Even if they did, you don’t let facts get in the way of a good story
But isn’t speed important? Imagine I ask 5 billionaires whether they eat breakfast. They all say yes. I conclude that eating breakfast will make you a billionaire! Except there are 7 billion other people who consume food in the morning while sitting on their (non-billionaire) asses. Every manager in my career wanted me to do things faster. None of them were ridiculously rich or named Jeff Bezos.
The genre
Most of the literature (the business kind especially) presents survivorship bias as cause and effect.
“The Millionaire Next Door” is a great example of bias masquerading as analysis (9k reviews on Amazon, 4.6 out of 5, millions of copies sold). The book identifies 7 common traits among rich people (e.g. living below your means) but never tests those traits on people who aren’t millionaires. Is living below your means actually rare among everyone else? Millions of people do nothing but. The book never checked, so the trait can’t do the work it’s being asked to do.
In “The Blue Zones” (1.9k reviews, 4.6 out of 5), the authors identify regions where people live longer than average and give reasons for the longevity (e.g. family ties and a busy social life). But aren’t social ties important in every society? And which way does the arrow point — would you have a busy social life if you were sick?
People recommended “Good to Great: Why Some Companies Make the Leap and Others Don’t.” The author identifies traits of great companies but never checks whether the same traits were present in the (now) bankrupt ones.
“I interviewed 800 successful startup founders and here’s what I found.” Useless. For every lucky founder there are thousands of unsuccessful ones nobody interviews. “Here’s one thing Buffett looks for in a company.” Stupid. Aren’t millions of non-Buffetts looking at the same metric? And none of them are billionaires.
Ten years of only winners
During my consulting years I analyzed vehicle launches for clients. Every client wanted to study the launches that performed exceptionally well. Only winners. Nobody cared about losers. Nobody ever asked me to take the insights from a successful launch and test them against the failures. I had that job for 10 years.
In my last year, one automotive company had a series of successful launches. The OEM was small and all of its vehicles had all-wheel drive (AWD) — a very salient feature. Another, less successful OEM zoomed in on this and started offering AWD vehicles in shrinking segments, claiming it had found a silver bullet. Millions of dollars were spent and sales kept going down. There were already AWD vehicles from other OEMs in those segments. But those cars were losers, so nobody bothered to check.
Which brings us to your portfolio
Every backtest runs on the tickers that still exist. Every “study the great compounders” framework is Good to Great with a Bloomberg terminal. Every screen I’ve ever built started from the names that worked. The delisted, the acquired, the bankrupt don’t show up in the sample. If your process was fit on survivors, you don’t have a process. You have a description of what already happened.
Unlike the business-book version, this one has actually been measured, and the bias is not small.
Malkiel looked at equity mutual fund returns from 1971 to 1991 and found that computing them from surviving funds alone overstated performance by roughly 1.5 percentage points a year. Elton, Gruber and Blake, running a risk-adjusted version over 1976–1993, put the bias at about 0.7 to 0.9 points. Carhart and co-authors found the distortion grows with the window: negligible over one year, past one percentage point annually once the sample runs longer than fifteen years.
Sit with the compounding. A one-point annual overstatement is a 10% overstatement of terminal value over ten years, 22% over twenty, 35% over thirty. At Malkiel’s 1.5 points, twenty years gets you to 35%. That entire gap is manufactured by funds that closed, merged, or were quietly folded into a better-performing sibling. Nobody lied. The dead funds simply stopped filing, and the database kept only the ones that were still around to be counted.
The same mechanism runs through everything else on the desk. Index membership studies drop the constituents that got removed. Factor research is built on data vendors whose coverage of dead securities is thin and whose coverage of dead securities in emerging markets is worse. Strategy backtests are almost always run on today’s universe, and today’s universe is defined by having survived to today.
When is it worth studying winners at all?
It depends on the distribution of winners.
Human lifespan is bounded. Nobody has made it past 123, and plenty of people reach 100. The gap between the best possible outcome and a merely good one is small, so the top of that distribution is crowded — and a crowded top is hard to reach on luck alone. Whatever the centenarians have in common is at least worth a look.
Wealth has no such ceiling. We had millionaires, then billionaires, and we’ll have trillionaires. Nothing caps the top, so the tail runs long and thin with very few people in it. The more skewed the distribution of winners and the smaller the sample, the bigger the role of luck.
This is the part that matters for a portfolio. “What did the ten-baggers have in common?” is a question about the thinnest, most luck-saturated part of an unbounded distribution, answered from a sample of a few dozen. “What did the companies that survived twenty years have in common?” is a question about a crowded, bounded outcome with thousands of observations. The second one can be answered. The first one mostly cannot, which is why the answers to it keep changing every few years and always describe the last cycle.
Since there are always more losers than winners, adding the losers increases the sample size and can flip the advice from positive to negative. “Eat your vegetables” becomes “don’t smoke,” because smokers drop out of the sample faster than people who skip vegetables. Nassim Taleb makes this point in Antifragile: charlatans are recognizable in that they give you positive advice, and only positive advice. Subtractive epistemology deserves its own post.
So: what do the most successful people do before breakfast? We will never know.
Sources for the fund figures: Burton Malkiel, “Returns from Investing in Equity Mutual Funds 1971 to 1991,” Journal of Finance (1995); Elton, Gruber & Blake, “Survivorship Bias and Mutual Fund Performance,” Review of Financial Studies (1996); Carhart, Carpenter, Lynch & Musto, “Mutual Fund Survivorship” (2002).