Survivorship bias is a cognitive bias that occurs when an analysis or decision-making process focuses on successful individuals or things while ignoring those that did not survive due to lack of visibility.
This bias can lead to overly optimistic beliefs and flawed conclusions because the failures are ignored.
The aircraft example
One of the classic examples of survivorship bias can be found in the context of World War II aircraft. During the war, engineers analyzed returning planes to determine which areas required reinforcement to increase their chances of survival.
Initially, they focused on the bullet holes and damage visible in the surviving aircraft. However, the critical insight came when a statistician named Abraham Wald suggested that they should reinforce the areas where returning planes showed little to no damage, as those planes might not have survived if they had sustained damage in those areas.
This insight illuminated the folly of only considering successful outcomes without acknowledging those who failed.
Survivorship bias in investing
In investing, survivorship bias can manifest when looking at the performance of mutual funds.
More often than not, only the funds that continue to operate report historical performance, while those that failed or were dissolved are excluded from performance studies.
Hence, newer investors may mistakenly believe that investing in mutual funds is a robust and consistently profitable endeavor, overlooking the many funds that underperformed to the point of being shut down.
How to reduce the bias
To mitigate the effects of survivorship bias, it is important to adopt a comprehensive view of available data. This includes explicitly considering the failures and the underlying reasons for those failures.
By acknowledging both successful and unsuccessful outcomes, it becomes possible to develop a more accurate understanding of the challenges and factors influencing success.
In conclusion, survivorship bias serves as a reminder of the importance of critical analysis and comprehensive data evaluation.