Survivorship Bias
Survivorship bias is the distortion that creeps into performance data when the failures are quietly removed from the sample. Funds that closed, stocks that delisted, strategies that blew up: none of them are around to be averaged, so the survivors’ record gets presented as if it were everyone’s record.
The math
Imagine 100 hypothetical funds launch in the same year. A decade later, 60 remain, having averaged 9 percent annually; the 40 that closed averaged 2 percent before shutting down.
The honest average across all 100 launches is 6.2 percent, yet every database now shows 9, because only survivors report.
| Survivors only | All 100 launches | |
|---|---|---|
| Funds counted | 60 | 100 |
| Average annual return | 9% | 6.2% |
| $100,000 after 10 years | ~$237,000 | ~$182,000 |
The $55,000 gap was never anyone’s return. It is an artifact of counting only the ships that came home.
The trap
Backtests built on today’s winners. Testing a strategy against the current membership of an index, or the stocks that exist today, silently excludes every company that went bankrupt or delisted along the way, and those are exactly the stocks the strategy would have owned and ridden down.
The same distortion inflates fund marketing, newsletter track records, and the anecdotes of investors at dinner parties: the losers stopped talking, so the sample skews cheerful.
The move
Ask “compared to what, counting whom” before trusting any performance claim. Serious research uses point-in-time data that includes dead companies and closed funds; anything less deserves a discount.
The bias cuts every direction, which is the useful part: it flatters star managers, and it equally flatters the tidy narratives around past index returns, which include no future and all the hindsight. A stock picker’s protection is structural, judging each business on its own forward economics rather than on any curated sample of past winners.