The "smoothing out" thesis
Some say that markets can be efficient
even if investors have cognitive and emotional biases.
A kind of collective self organization would smooth out deviations.
There would be two reasons for that:
1) Investors have different risk attitudes and different risk / return / price estimates, more
or less close or far from the "fair price".
But, in very liquid markets with many buyers and sellers, those estimates are supposed to
obey a Gaussian distribution (=bell shaped) curve.
There would be in that case:
As many risk-prone investors on one side of the risk attitude curve, than there
are risk-averse ones on the other side.
As many investors who underestimate the right value than investors who overestimate
it, on each side of the median / the mean of the distribution curve of estimated prices.
With a cluster near the price median / mean, and an harmonious and symmetric
distribution around it.
=> Thus the market price is supposed to be equal to that median / mean, and to reflect the
best estimate of the real value.
2) even if that was not the case, and if the market price became too high or too low
compared to the "real value", the most rational investors will seize the opportunity.
They would sell or buy until the price correspond to the fair one.
They will intervene quickly, in a race, so as not to miss that "arbitrage" opportunity.
=> Thus, very soon, the "right" price is reached.
Or at least the quotations would undulate nicely around it and quite close to it, with a small
There would not be any worthy arbitrage opportunity left.
Two answers to that thesis
What if errors compound instead of compensate?
1) Nothing proves that there is an harmonious Gaussian distribution of
investors attitudes or estimates.
Mathematical studies show that returns, prices and volatilities muster, on a chart, clusters,
skews, kurtosis, fat tails, flatnesses, dents and bumps and other anomalies compared with
the "normal" (Laplace-Gauss) law. Or in some cases, power laws (Paretian distributions
with concentrations on the extreme data)
One reason is that some investor classes with more money than others tend to dominate
the market. Nothing prove that their estimate is not biased even if they are reputed to be
better informed and more "professional".
But the main reason is collective biases, where people tends to imitate one another in
underreacting / overreacting.
As long as the mass of investors acting in a given biased direction is too small, their
actions get canceled out and the market might stay efficient.
But when that mass gets sizeable enough, it reaches a *critical point* where more
and more other investors are attracted. Individual behaviors turn into collective
This triggers a cascade / chain-reaction / snowball effect. There is a growing
skew in the distribution curve of expected prices and, more important, a deviation
of its mean from the "fair price"
2) "Rational" investor might consider that to calculate overpricing or underpricing is not relevant
as long as price obeys a trend and that this trend is not reverted.
=> So their interest is to play on such trends as "conscious followers" (rational expectation theory)
instead of playing against them.
Such a behavior adds to the momentum (positive feedback) instead of moderating it (what a
negative feedback or reversion to the mean would do).
Some influent investors (big hands or opinion leaders) can even have a tendency to "manipulate"
the market that way.
Well, maybe we should identify better
Of course, if we define EMH as perfect reaction to information, those rational or irrational
exaggerations can be interpreted as a specific version of the EMH. We could label it the
"BF-enlarged EMH" (*).
Prices are driven by new information, as the EMH states. But many investors might give more
importance to some information than to others.
In other words the pieces of news
that come from the asset market :
In other words those on economic
fundamentals: economic risks and
uncertainties... earnings, interest
rates, growth prospects,
(*) see also Martin Sewell's paper on a "behavioral-based RWH").
This page fist published Jan 29th, 2003. Last update: