Scott Sumner succinctly explains how illusory bubbles can appear in a world ruled by the Efficient Markets Hypothesis:

How should Bitcoin be priced?  If there is a 95% chance that it will
soon be worthless and a 5% chance that it will soon hit $1000, then $30
seems like a relatively fair price.  That allows for a substantial
expected gain ($50 minus interest costs would be the risk-neutral
price.)  But Bitcoin is very risky, so investors need to be compensated
with an above average expected rate of return.

Now consider a point in time where the asset is selling at $30, and
investors have not yet discovered whether it will eventually reach
$1000.  Should you predict that the price is a bubble?  Yes and no.  It
is likely to eventually look like it was a bubble at $30.
 Indeed 95% of such assets will eventually see their price collapse.
 That’s “statistically significant.”  It’s also significant in a
sociological sense.  Those that call “bubble” when the price is at $30
will be right 95% of the time, and hence will be seen as having the
“correct model” of bubbles by the vast majority of people.  Those who
denied bubble will be wrong 95% of the time, and will be seen as being
hopelessly naive by the average person.  And this is despite the fact that in all these cases there is no bubble, as by construction I assumed the EMH was exactly true.


I predict that eventually the price of Bitcoins will fall sharply
(from some level of which I am not able to predict) and people will
vaguely recall:

“Wasn’t Scott Sumner the guy who denied Bitcoins was a bubble?  What an idiot.”

Of course, this story is a lot easier to believe for weird new assets like Bitcoin than for familiar old assets like the entire U.S. housing stock – especially when investors actually wrote down their probability distributions.