Survivorship Bias: The Records We Never See

*Every account of investing you encounter was written by someone who was still there to write it. The ones who were not are not silent by choice.*

There is a distortion that affects nearly every piece of evidence an investor encounters, and it operates so quietly that most people never notice it. The distortion is that the failures are missing. Funds that performed badly are closed and disappear from the databases. Companies that collapsed are removed from the indices. Investors who were ruined do not write books about their methods. What remains, and what is therefore examined, is the portion that survived, and the survivors are not a representative sample of what was attempted.

The consequence is a systematic overestimation of how well things go. Consider the record of a category of investment funds over twenty years. The figure is typically computed from the funds that exist today and have twenty-year records. But funds that performed poorly were closed or merged into others during that period, and their records vanished with them. The surviving funds are, by construction, the ones that did well enough to survive, and the average of their records is therefore substantially better than the average experience of someone who invested in that category twenty years ago and did not know which funds would endure.

Studies attempting to correct for this have generally found that the effect is not trivial. When the closed funds are added back into the calculation, the average performance of the category falls meaningfully, and the proportion of funds that outperformed a simple market holding falls further still. An investor examining the uncorrected figures is looking at a picture from which the disappointments have been quietly removed, and they are drawing conclusions about the odds from a sample that excludes most of the bad outcomes.

The same distortion applies to individual companies and to indices. A market index today contains the companies that have prospered enough to remain in it, and its long-run record reflects the periodic removal of those that failed and their replacement by those that succeeded. This is not a defect in the index, which is doing precisely what it is designed to do, but an investor who supposes that the index's record represents what would have happened to someone who bought its constituents decades ago and simply held them has misunderstood what they are looking at.

It applies with particular force to the accounts of successful investors, which are the most widely read literature in the field. The investors who succeeded have written extensively about their methods, and their books are examined for the principles that produced their results. But a great many investors employed similar methods and did not succeed, and they did not write books, because failure is not something people publish. The apparent evidence that a method works is therefore drawn entirely from a sample of people selected on the basis of having done well, which is precisely the selection that makes the evidence useless for establishing whether the method works.

This does not mean that skill does not exist or that nothing can be learned from those who have succeeded, which would be an overcorrection into cynicism. It means that the number of people who employed a given approach and failed is invisible, and without that number the success rate cannot be calculated. An approach that produced ten spectacular successes might have been employed by twelve people or by twelve thousand, and these two possibilities imply entirely different conclusions about whether an ordinary investor should adopt it.

The bias is compounded by a related distortion in how stories are selected for retelling. The narratives that circulate are the dramatic ones: the investor who concentrated everything in a single company and became wealthy, the early participant in a transformative technology. These stories are memorable precisely because they are unusual, and they are unusual because most people who attempted the same thing did not achieve the same result. The rarity that makes a story worth telling is exactly the rarity that makes it a poor basis for a decision.

The practical defence begins with a habit of asking, of any evidence, what has been removed from it. When examining a record of returns, one might ask what happened to the participants who are not in the sample. When reading an account of a successful method, one might ask how many people employed something similar and are not being discussed. These questions rarely have precise answers, and that is itself informative: the absence of an answer indicates the size of the hole in the evidence.

There is a personal application of the bias that deserves mention, because investors commit it against themselves. An investor reviewing their own history will tend to recall the decisions that worked and to explain away those that did not, and this is survivorship bias operating within a single memory. The written record recommended throughout this project is, among other things, a defence against this: it preserves the decisions that failed, in their original reasoning, and prevents them from being quietly edited out of one's own account of one's own competence.

The bias also explains why the historical episodes examined in this section are so easily misread. The bubbles that burst are studied; the manias that quietly deflated without drama are not. The frauds that were exposed are famous; those that were never detected are, by definition, absent from every list. An investor forming their sense of how markets behave from the episodes that were dramatic enough to be recorded has again been shown a selected sample.

At VESTFY™ survivorship bias is presented last among the behavioural cases because it governs how all the others are perceived. Every lesson an investor draws from the record is drawn from what remained, and what remained was chosen by the very process one is attempting to understand. The appropriate response is not paralysis but a persistent, slightly uncomfortable question, asked of every impressive record encountered: who else tried this, and where are they now?