Survivorship bias
FlashcardsQuick reviseKey facts. One by one.Start →Topic: Economic Data: Census, NSS, Surveys and Statistical Tools · NCERT: Beyond NCERT
Meaning
Survivorship bias means judging a method, a group or a policy only by the cases that "survived" or succeeded. The cases that failed or dropped out are left out of the count. The survivors are usually the better performers, so the method looks more effective than it really is. In statistics, it is a kind of sampling bias: the people or firms who failed are missing from the list we study, so the sample does not stand for the whole group. It is a non-sampling error, so collecting a bigger sample of survivors does not fix it.
Example
An analyst wants the average return of mutual fund schemes over ten years. They study only the schemes that still exist today. Schemes that did badly and were closed or merged are not on the list. The average therefore comes out too high, and a saver may think fund investing is safer than it really is. The same mistake happens when a talent programme is praised for one champion while nobody counts the many trainees who dropped out under that same programme.
Don't confuse with
- Sampling error: the gap between the true population value and the sample estimate that comes from studying only a part of the population. A larger random sample reduces it. Survivorship bias comes from a frame that leaves out the failures, and a larger sample does not reduce it.
- Non-response error: the sampled units are on the list but cannot be reached or refuse to answer. Under survivorship bias, the failed units were never on the list at all.