Efficiency in Markets
Scott Sumner calls in the EMH opposition
I think efficient market hypothesis (EMH) makes a whole lot of sense. Large financial markets with many participants should, and do, tend not to leave unexploited opportunities for gains, whether from new information or through arbitrage. The empirical evidence for efficient markets at the aggregate level, and over a longer time frame, is also simply overwhelming as well. (See here for a collection of literature.)
Yet there are limits. And Scott Sumner seems not to see them.
Paul Krugman, in one of his more "I-told-you-so" political screed pieces for The New York Times during the worst of the recession, wrote that behavioral economics goes along way to explaining why markets can misprice assets. I agree with Krugman's contention here, despite his bombast: irrational panic and exuberance exists in markets and is generally standard behavior when you examine any crowd in constant communication. You can get very accurate predictions of probability, because of the decentralized calculations all of the market participants are making (witness Intrade). But that group loses its decentralized nature when relevant information is added, particularly when it is contradictory and uncertain. The group begins to coordinate its decisions, as Keynes famously wrote in his characterization of stock picking as a "beauty contest," to avoid taking a loss, even if they knew they were right.
Looking at major repricing events in large markets, we can see that the evidence for some sort of crowd-behavior effect in markets in not easily dismissed, but that markets often reach efficiency sooner than later.
But here's another interesting question: since EMH assumes that the interests of market efficiency and those of the individual investor are always in perfect alignment, it must presume that the costs for investors to make decisions are zero. (Otherwise, investors wouldn't take action to address mispricings.) There are rather dry explanations for why costs are not zero: transaction costs for stock brockers, banks, etc.
These are fairly small, though--they won't explain the billion-dollar-bills that seem to lie in the street every now in then. So now we add in the crowd effect more formally into our model: assume that there is a cost for an individual investor to sound a contrary note. This may come in the form of being fired or questioned at work; I'm thinking about contrarian analysts, and how they get some frankly rough treatment. Also, if an investor runs against the grain, then they are probably expecting to take a short-term relative loss, which suggests that they must apply some sort of discount rate to future gains different from other investors. So to the extent that this discount rate makes contrarianism unprofitable, the market doesn't have to be instantaneously inefficient.
But wait, that's not all. If markets are perfectly efficient, and assets are priced properly given all available information, who's keeping the asset prices in perfect alignment? Certainly not investors, because arguably there is no opportunity for gain. A common criticism of EMH is that the market simply doesn't look at all as it would predict: messy, busy, volatile--not smooth, mechanical, or low-volume. I'm trying to take this EMH-is-not-everything a bit further, though: it feels to me like an undiluted assumption from the classical model...not like a respectable position in an era where we tend to make some Keynesian assumptions in macroeconomics.
So when we make prices stickier, and say that investors are human and must be compensated, in effect, by the market, for acting on new information and taking a risk, EMH now takes on a much more nuanced, New Keynesian tenor.