Vicarious Learning, Undersampling of Failure, and the Myths of Management
Risky practices appear positively correlated with success in survivor samples even when unrelated to performance
This is the definitive paper on survivorship bias in business research and management theory. Denrell demonstrates mathematically that risky practices - even if completely unrelated to performance - will appear positively correlated with success when we only study survivors. This explains why business case studies systematically recommend concentrated resource allocation and other risky practices that the full population data would not support.
The paper is foundational to the book's argument that biology provides better data than business. Biologists study the full population of organisms, including failures. Business researchers typically study only successful companies, learning the wrong lessons from an unrepresentative sample. A principle derived from fern drought resistance is more reliable than a case study of what worked for one company.
Key Findings from Denrell (2003)
- Risky practices appear positively correlated with success in survivor samples even when unrelated to performance
- Business case studies systematically overestimate the value of concentrated resource allocation
- Survivorship bias is a mathematical inevitability, not just an occasional error
- Management 'best practices' often reflect sampling bias rather than causal relationships