Survivorship Bias in Historical Value Strategy Tests

Summary

The document asks whether a historical value-strategy test beginning after a severe market decline might omit companies that failed or left an index during the downturn. The proposed study ranks stocks annually by price-to-cash-flow ratio, assigns them to five portfolios, and evaluates performance over a later period. It recognizes that each holding year should use the index membership known at portfolio formation rather than today’s constituents.

The post poses the question but gives no answer or empirical evidence. Its key methodological concern is distinct from using period-appropriate constituents: starting the sample after a crash can still leave a researcher without firms that disappeared before the first observation. A valid test needs point-in-time membership and return data that include delisted and failed companies where relevant. Conclusions will also depend on the chosen universe and accounting-data timing, details not supplied in the document.

Key ideas

• Annual portfolio formation should use the constituent universe available at the formation date.
• A study beginning after a market crash may omit firms that failed or disappeared before the sample starts.
• Survivorship-aware testing requires historical membership and returns that include delisted firms where relevant.
• The document asks the question but supplies no empirical resolution.

Tags

# Possible survivorship bias


# Possible survivorship bias












I am quite new to studying finance so apologies if this turns out to be a trivial question. I am writing my first essay where I want to test a value strategy from 2009-2019 by yearly ranking of the stocks by their P/CF ratio, forming 5 portfolios, evaluating their performance.

Regarding survivorship bias, I understand that it is necessary to consider the members of the index in each year under analysis (i.e. for the holding year 2009 - 2010, the index constituents per 2009 are to be considered) But I am unsure if 2009 as a starting year would cause issues, meaning that only the survivors of the market crash are considered. Do you see any issue with this or am I making a thinking error?

Thank you for your help.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

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