Why do "average returns" often look too good?

A tempting study: "the average S&P 500 company returned X% over 30 years — just buy good companies!" The math may be flawless and the conclusion still broken, because of the quietest force in financial data: the losers vanish from the lists.

Survivorship bias

Compute returns from TODAY'S index members backward and you've silently excluded everyone who didn't survive the journey: the bankrupted, the delisted, the acquired-in-distress. Course 2 taught you the index list changes constantly — companies get removed for failure routinely. History computed on survivors is history with the funerals edited out; it MUST look rosier than what an investor living through it experienced.

The same ghost haunts fund statistics (bad funds get closed and disappear from averages), strategy folklore ("everyone I know who bought X got rich"; the ones who lost stopped talking about it), and — critically for your Quant path — any backtest run on a "current constituents" list.

Survivorship's siblings

• Look-ahead bias: using information at a date before it was knowable — computing with full-year revenue in January, trading on an earnings figure the night before it printed. Time-machine math.
• Cherry-picked windows: "this strategy tripled!" — measured start-of-2009 (the bottom) to end-of-2021 (the top). Same strategy, honest window, different story.
• Backfill: a fund joins a database and brings its glowing pre-inclusion track record along; the databases fill with historical winners retroactively.

The one question that catches them all

"Who is missing from this sample, and could they have been known at the time?" Ask it of every impressive average, every strategy pitch, every "study shows." Most collapse politely. This single question is among the highest-value tools this Academy will ever hand you — and it costs nothing but the asking.

In the data

Market data keeps the living and the dead on separate lists. The list of what trades on an exchange shows the living by default; the companies that left sit on a delisted list of their own. Here are the first three on London's:

Ordinary names, most of them — a company leaves the list for a merger, a buy-out or a move as readily as for a collapse, and a study built only on the living list never sees any of them. Indices work the same way: the fix for survivorship is a dated membership history, every company ever in the index with the day it joined and the day it left, filtered to the date you are measuring — funerals included.

Try it now

1. Count the graveyard. We counted both of London's lists on 29 September 2026: 7,270 instruments trading, and 4,080 delisted. Divide the second by the first. The answer is more than half — thousands of companies that a study of "current constituents" would never see, and every one of them is a story that ended. For scale, London's own record is below.

2. Name two or three famous companies that left major indices in disgrace. Enron and Lehman Brothers are the two everyone remembers; add any fallen giant of past decades you know of.
3. Note that today's "index average since 1995" contains none of their pain, and that you have just measured roughly how many funerals that is.
4. Next time you meet an amazing average, ask the question out loud. Feel the study squirm.