How survivorship bias distorts reality
When I purchased my last car I was amazed to find that the license number I was given was 6NWL485. What are the chances that I will get this combination of letters and numbers? Before I got the number, the odds were 1 in 175,760,000. (The number of letters in the Latin alphabet to the power of the number of letters that appear in the number times the number of digits from 0 to 9 to the power of the number of digits that appear in the license number: 26 in the third times 10 in the fourth) However, after I received the number, the probability is 1.
Economist Gary Smith of Pemona College calls this phenomenon "survivorship bias." In his insightful book "Standard Deviations" (Overlook Publishing, 2014) he cites this bias as an example of one of the many perceptual biases associated with statistics. Smith demonstrates it using a series of cards: 3 of clubs, 8 of clubs, 8 of diamonds, queen of hearts and ace of hearts. The chance of getting that combination is about 1 in three million, but, as Smith says, "Having already received this suit and looking at the cards, the probability of holding those five cards is 1, not 1 in 3 million."
When you think about it, the conclusion seems obvious. But most of us tend to fall into the trap of survival bias. Take for example the multitude of business books sold at every bookstore and dealing with the most successful commercial companies. Smith analyzes two bestsellers of this genre. Jim Collins, author of the 2001 book Good to Great (which sold more than three million copies), selected 11 commercial companies, out of 1,435, whose stocks outperformed the market average over 40 years. He then researched and looked for common characteristics that he believed might explain the success of these companies. Instead, Smith says, Collins should have started with a list of companies at the beginning of the review period and then used "reasonable metrics that would allow him to predict which 11 companies will do better than others. You have to apply those metrics objectively, without looking at what the companies will achieve over the next 40 years." It is not fair, and it has no meaning, to predict which companies will succeed after checking which ones have already succeeded! This is not a prediction, it is history." In fact, Smith adds and writes, from 2001 to 2012 the shares of six of Collins' 11 "big" companies did less well than the market average. This indicates that his method for retrospective analysis is fundamentally flawed.
Smith encountered a similar problem in the 1982 book In Search of Excellence (which sold more than three million copies), in which the authors, Tom Peters and Robert Waterman, found eight common characteristics of 43 "excellent" companies. Since then, Smith writes, of the 35 companies on the list whose shares are traded on the stock exchange, the achievements of 20 have been below average.
Survivorship bias was also prominent in the reception of another 2011 bestseller, Walter Isaacson's biography of Steve Jobs. Readers tried their best to understand what made this volatile genius so successful. Do you want to be the next Steve Jobs and start the next Apple company? Drop out of college and start a business in partnership with your friends in your parents' garage. How many people have done this and failed? who knows No one writes books about them and their failed companies. But venture capital funds have data on the probability that the start-up company founded in a private garage will succeed "big time". And here the survival bias is of a different kind.
And so David Kwan of the Bessemer Capital Fund in Menlo Park, California, wrote to me in an e-mail: "In order for entrepreneurs working in their home garage to reach the top percentile in America, they must in most cases go through a route that includes raising venture capital and then issue their startup on the stock market or be acquired by another company If their garage is in Silicon Valley, they will have to submit proposals to about 15 venture capital funds that each such fund examines about 200 different proposals. Therefore, maybe one out of every 13 startups gets funding. Even after that, it is still a long way off On its website, 2013 was a typical year: out of the 1,334 start-up companies that received funding, only 13% went public (81 companies) or were purchased for an amount high enough to require a public announcement (95 companies). That is, for every rich startup founder, there are 100 other entrepreneurs left with a messy garage."
Surviving this statistical chance is truly a rare event.
About the author
Michael Shermer is the publisher of Skeptic magazine (www.skeptic.com) His next book is: "Science's Moral Noah's Ark". Follow him on Twitter: @michaelshermer
The article was published with the permission of Scientific American Israel
2 תגובות
My friend,
Where does your extreme confidence come from? Shouldn't you stop for a moment and question it?
Have you heard of conditional probability?
Probability given that an event has happened is 1 regardless of the unconditional probability that it will happen
There is a problem with this whole argument. And in my opinion the author reached the opposite conclusion to the correct one.
Probability is a number that we attach to a certain event. This number is calculated mathematically and therefore cannot change depending on what happens or does not happen. This number is above the events themselves. The fact that he received a certain license number did not change his probability of receiving it. So either the probability was always 1. Or it was always 1 divided by 175 million.
In short, you can write an identical article only in reverse: "How events distort the perception of probability"