A philosopher of science (Duke’s Alexander Rosenberg) does an awfully good job of explaining the core problem with the efficient market hypothesis (via Zubin Jelveh):
The efficient markets thesis is that the market makes complete use of all relevant information, and the “proof” is roughly that in a perfectly competitive market among perfectly rational agents prices invariably and instantaneously reflects all agents’ real beliefs and real desires. Any one who knows anything that can make him or her money acts on it—buys or sells—and that signal is picked up by every one else, who also acts on it, thus preventing any one from making excess profits—rents—long-term.
The first thing a philosopher notes about this notion is that since most people have false beliefs, especially about the future, an efficient market doesn’t internalize knowledge, but only beliefs. If they are mostly false, then the market isn’t efficient at internalizing (correct) information, it’s efficient at internalizing mostly false beliefs. If false beliefs are normally distributed around the truth, then they’ll cancel out and the proof of a probabilistic version of the efficient markets theorem will go through—market prices reflect the truth most of the time. Too bad false beliefs don’t always take on this tractable distribution.
I make this case in my book, but not as succinctly. I’m also a little appalled that I had never heard of Rosenberg or his book Economics: Mathematical Politics or Science of Diminishing Returns? In an efficient market of ideas I would have already read it.
Update: An even more succinct explanation, from the original 1934 edition of Graham and Dodd’s Security Analysis:
[T]he market is not a weighing machine, on which the value of each issue is recorded by an exact and impersonal mechanism, in accordance with its specific qualities. Rather should we say that the market is a voting machine, whereon countless individuals register choices which are the product partly of reason and partly of emotion.