Definition
Survivorship bias arises when the sample contains entities that remain observable while entities that disappeared, failed or were delisted are omitted. The surviving group can then make historical outcomes look better or more stable than the full opportunity set. A point-in-time universe and delisting treatment are needed to address it.
How it is calculated
The mean outcome among survivors S=1 can differ from the mean across both survivors and non-survivors.
| Symbol | Meaning | Unit |
|---|---|---|
| E | expectation | outcome units |
| Y | outcome | return or performance units |
| S | survival indicator | binary |
Interpretation
Looking only at companies still standing is like judging a tournament from the finalists and forgetting everyone eliminated. The survivors are real, but they are not the historical universe. The omission can distort averages, rankings and risk estimates.
Assumptions and conventions
-
The target universe includes entities that may leave the sample.
-
Entry, exit and delisting outcomes are measured consistently.
Outcome units; S is dimensionless.
Limitations and common misreadings
-
Survivorship bias is one form of selection bias, not a synonym for all selection bias.
-
Adding more surviving observations does not restore omitted failures.
References
- Stephen J. Brown, William Goetzmann, Roger G. Ibbotson, Stephen A. Ross. (1992). Survivorship Bias in Performance Studies. Review of Financial Studies. Whole article. doi:10.1093/rfs/5.4.553. Source

