Survivorship Bias in Trading Research and Backtests

Summary

Survivorship bias arises when analysis includes only assets or funds that remain observable, omitting those that failed, delisted, or closed. The document explains the issue with the military aircraft example: examining only planes that returned can misidentify where protection is most needed. In investment research, dropping weak or discontinued holdings can make historical portfolio results look too favorable, while also distorting the apparent performance of a market or strategy.

The article illustrates the effect with a mutual-fund comparison: the average return across all funds is stated as 1.94%, versus 3.58% for the three best performers. It recommends using data sources that retain investments no longer in existence and including poor performers in portfolio valuation and backtests. The examples make the selection problem clear, though they do not quantify its prevalence across markets or offer a complete procedure for correcting other forms of bias. Reliable conclusions still depend on representative data and careful study design.

Key ideas

• Survivorship bias occurs when failed or discontinued assets are omitted from analysis.
• Backtests using only currently available securities can overstate historical results.
• The document's mutual-fund example shows how selecting top performers raises the reported average relative to including all funds.
• Use datasets that retain delisted securities and closed funds to reduce this bias.
• Avoiding survivorship bias does not by itself correct other weaknesses in a backtest.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.