Week 35 Day 3: Survivorship Bias: The Hidden Graveyard of Failed Investments

A fund company markets its '20-Year Track Record Fund' showing 12% annual returns. Impressive! But over those 20 years, the company launched 50 funds. 30 of them closed because they performed poorly. You are only seeing the 20 that survived. If you include the closed funds, the average return drops to 6%. The survivors look great only because the failures are buried.

Survivorship bias everywhere in finance: (1) Mutual funds. Approximately 40% of mutual funds that existed 20 years ago have since closed or merged (usually due to poor performance). The remaining funds' track record is biased upward. Morningstar estimates survivorship bias adds 1-2% to apparent category returns. (2) Hedge funds. The Hedge Fund Research Database has a massive survivorship bias: funds that blow up simply stop reporting. The 'hedge fund performance' you see in the news includes only the survivors. When accounting for dead funds, hedge fund alpha drops to approximately zero. (3) Stock market indices. The S&P 500 constantly adds strong companies and removes weak ones. Sears, Kodak, and Lehman Brothers were all S&P 500 members -- until they were not. The index always looks healthy because sick members are replaced. (4) Success stories. You hear about the person who invested $10,000 in Amazon in 1997 and has $20 million. You do not hear about the thousands who invested in Pets.com, Webvan, and e-Toys -- companies that went to zero. (5) Real estate. 'Nobody ever lost money in real estate' ignores every foreclosure, every underwater mortgage, and every developer who went bankrupt. How to correct: whenever you see a performance claim, ask 'What is not being shown?' What funds closed? What companies failed? What strategies blew up? The full picture includes the graveyard, not just the survivors.

Survivorship bias was formally addressed in finance by Elton, Gruber, and Blake (1996), who estimated that survivorship bias in mutual fund databases inflates average fund returns by approximately 0.7-1.5% per year -- a significant portion of the apparent 'alpha' claimed by active managers. Brown, Goetzmann, Ibbotson, and Ross (1992) showed that survivorship bias in hedge fund databases is even more severe: approximately 3-5% per year, because hedge funds that fail stop reporting voluntarily (there is no requirement to report to databases), and databases retroactively delete the records of failed funds. Fung and Hsieh (2000) additionally identified 'backfill bias' -- when a newly reporting fund adds its historical (audited) track record to a database, only funds with good track records choose to report, creating an additional upward bias. The combined effect of survivorship bias, backfill bias, and selection bias (only funds that believe they have alpha choose to report) means that the publicly available hedge fund return data overstates true hedge fund performance by approximately 3-7% annually. After these corrections, the aggregate hedge fund industry delivers approximately zero alpha -- consistent with the efficient market hypothesis. The broader epistemological lesson: survivorship bias is a form of selection bias (the subset of observations you see is non-randomly selected from the full population), and it is pervasive in precisely the domains where people make the highest-stakes decisions. The correction requires deliberate, effortful inclusion of base-rate data (how many funds launched? how many survived?) -- data that is almost never presented alongside success stories.