Survivorship Bias — The Track Records That Disappear
Poorly performing funds tend to quietly disappear from the record — flattering the industry's own reported average.
A specific, often underappreciated distortion in how mutual fund performance gets reported: funds that perform poorly for long enough are disproportionately likely to be closed, merged into another fund, or otherwise quietly removed from the very databases used to calculate "average" fund performance — meaning the historical record available today is biased toward the funds that happened to survive.
This means a simple average of "all funds that exist today, going back 20 years" systematically overstates how a real investor, invested in the full original population of funds (including the ones that later disappeared), would actually have fared — the failures have been selectively edited out of the very data being used to judge the industry.
Bogle treats this as one of the more important, and more commonly overlooked, reasons that reported industry-wide fund performance looks better in retrospect than what investors, as a whole, actually experienced in real time.
The distortion compounds with the previous chapter's lesson about lucky short-term track records rather than replacing it. A fund that happened to top the rankings for a few years through ordinary statistical variation is more likely to survive long enough to build a longer track record in the first place, while a fund that had an equally unlucky stretch is more likely to be shut down before it ever gets the chance to revert back toward average — meaning the surviving population is doubly skewed, both by which funds happened to look good recently and by which funds were allowed to keep existing at all.
Suppose 100 similar funds launch in the same year. A decade later, 30 of them have underperformed badly enough to be closed or merged away, and only the 70 survivors remain in the industry's performance databases.
The industry's reported "average 10-year return for funds in this category" is then calculated using only those 70 names, since the 30 that disappeared simply aren't in the database being averaged. An investor glancing at that headline figure today has no easy way to know that nearly a third of the original population is silently missing from it — the number looks like a complete picture of the category's performance, when it's actually a picture of only the subset that happened to survive.
Calculating the "average 10-year return" using only those 70 survivors necessarily overstates what an investor who'd spread money evenly across the original 100 funds would actually have earned, since the 30 worst outcomes have been removed from the average entirely.
Survivorship bias is a much smaller issue for a fund that simply holds the whole market and doesn't get closed for underperforming its own benchmark, since there's no comparable culling of "bad" broad-market index funds relative to the market itself — the distortion specifically affects the reported track record of the actively-managed fund industry as a whole.
This gives index funds a further, less obvious advantage beyond the cost argument covered elsewhere in this course: their reported historical performance is a much closer, more honest reflection of what an investor who bought and held one would actually have experienced, precisely because there's no comparable population of failed, since-deleted index funds quietly distorting the record upward.
Survivorship bias doesn't announce itself the way an obviously misleading statistic might — the reported figure is, in a narrow technical sense, accurate for the funds it actually includes. The distortion is entirely in what's silently excluded, which means a reasonably careful reader comparing today's fund performance rankings can still be misled, not because the math is wrong, but because the population being measured has already been filtered by which funds survived long enough to be measured at all.
- Poorly performing funds are disproportionately closed or merged away, removing their results from the databases used to calculate industry-wide average performance.
- This systematically flatters reported historical fund performance relative to what a real investor, spread across the full original population of funds, actually experienced.
- The bias is invisible in the reported number itself — it lives entirely in what's silently excluded from the population being averaged.
- The effect is far smaller for broad index funds, which aren't culled the same way actively managed funds are, giving their reported track records more credibility.
- Reported "average fund performance" figures should be read with this bias in mind, not taken at face value.