There is perhaps no concept in economics that is as misunderstood and mischaracterized as the efficient markets hypothesis (EMH). A lot of people think they know what it means, but do not. A lot people think that they have clever arguments to refute it, but they do not. Today, I would like to discuss what the EMH says and does not say and clarify some arguments.
Eugene Fama’s original discussion and definition of efficient financial markets is as follows. Suppose that I want to try to determine the price of a stock tomorrow. I would use all of the information I have available to me to form some expectation today of what the stock price will be tomorrow. In an efficient market, the difference between this conditional expectation and the actual price would be entirely random and, on average, equal to zero. Put differently, what this really means is that no investor can systematically beat the market.
Unfortunately, many people took the wrong lessons from Fama’s presentation of the idea. I think that the wrong-headed criticism of the EMH stems from three or four main (and interrelated) misunderstandings. I would like to discuss each of these systematically.
The Random Walk Diversion and the Joint Hypothesis Problem
When Fama wrote his paper, he not only defined what he meant by efficient financial markets, but he also tested this hypothesis. One way to test Fama’s specific statement of the EMH is to ask what is meant by a conditional expectation of the future price of the stock. The simplest answer to this question is that a conditional expectation is just what some model of stock prices would predict the stock price to be, given the available information.
With that as a backdrop, consider a world in which information is costless. Everyone has access to all of the available information at any given moment in time. If this is true, then all of the relevant information that pertains to a particular company will be incorporated into its stock price. However, if all of the information that is available is already introduced in the stock price, then the only reason that stock prices could change from one day to the next is if new information comes available. Thus, it follows that the best prediction of tomorrow’s stock price is just today’s stock price.
This idea is what is known as the random walk hypothesis. It essentially says that changes in stock prices from one day to the next are random and that one’s expected value for the stock price tomorrow should just be the stock price today.
Fama used this as his jumping off point and found evidence to support the random walk hypothesis. Subsequent work seemed to confirm these results. Unfortunately, however, these initial results apparently convinced some people that the EMH was synonymous with the random walk hypothesis. Later, when others came along and found evidence against the random walk hypothesis, this was seen as a refutation of not just the random walk hypothesis, but also the EMH. In reality, this evidence against the random walk hypothesis did not refute the EMH. To understand why, we need to discuss a second misunderstanding: the joint hypothesis problem.
According to Fama’s original statement of the EMH, testing the EMH amounts to comparing a conditional expectation of the price to the actual, realized price. However, a conditional expectation of a price is just a model-based price. In other words, it is the price that a model predicts given the information that is included in the model.
Using this methodology, one takes a model-based prediction of the price and compares it with the actual realized price to see if the EMH holds. But this test really only works in one direction. If one finds that, on average, this conditional expectation is correct and deviations are random and unpredictable, this seems to provide evidence in favor of the EMH. However, if a researcher doesn’t get this result, it does not necessarily refute the EMH. The reason is that Fama’s test is a joint hypothesis test. When one uses Fama’s methodology, what they are doing is testing the joint hypothesis that financial markets are efficient and that the model used for comparison to the actual price is correct. In other words, if one uses Fama’s methodology, one cannot distinguish between whether the EMH is wrong or their model is wrong.
“Informational Efficiency” Versus “Economic Efficiency”
Naturally, the EMH calls out for theory. Unfortunately, theories designed to examine the EMH are often times asking a different question than Fama. In addition, there are a number of misconceptions that the theoretical literature helps to perpetuate.
Fama describes efficient markets as fully reflecting available information. In response to this claim, many people often cite a paper by Sanford Grossman and Joe Stiglitz that argues that it is impossible for markets to be informationally efficient. But I am not sure that this paper says what many people seem to think it says about the EMH.
Grossman and Stiglitz make an important point. In the real world, information is costly. In a world in which information is costly, an asset price cannot possibly reflect all available information. The intuition is quite straightforward, if information is costly, then people will invest in information up until the point at which the marginal benefit is equal to the marginal cost. In the context of financial markets, the marginal benefit of acquiring information is the additional returns that one can earn from that information. Since information is costly, people will acquire information until these additional returns are equal to the marginal cost. Of course, this necessarily implies that not all of the available information will be used and, as a result, prices will not fully incorporate every bit of information possible. In fact, if information was costless, this creates a conundrum. Since people set the marginal benefit of information equal to marginal cost, costless information implies that the marginal benefit is zero. However, if the marginal benefit is zero, there is no point to participating in the market and an equilibrium would not exist.
This conclusion seems like it refutes Fama’s argument that prices fully reflect available information, and perhaps it does in a certain way. How one interprets this claim depends on what one thinks Fama means by saying that markets fully reflect available information. Does this mean prices incorporate all information or does it mean that price incorporate all information for which the marginal benefit is greater than or equal to the marginal cost?
Regardless, the relevant question is whether or not this refutes the EMH, which is actually a separate question. The EMH states that one cannot systematically beat the market and this can be true even if all information is not reflected in the price. If there is information out there that would allow one to beat the market, but the marginal cost of the information exceeds the marginal benefit, then that extra information that is out there does not help one to beat the market.
Another avenue with which theorists have explored the idea of efficient financial markets is to (a) interpret the EMH as saying that price equals fundamental value, and (b) write down models of asset prices to determine whether they end up equal to some known fundamental value in equilibrium. While these models are no doubt valuable for helping us to understand how these markets work and what sorts of frictions can gum up the system, they are also missing the point.
Saying that financial markets are efficient when price equals fundamental value is the sort of thing that makes one sound smart to the uninitiated. Unfortunately, there is this pesky problem of what is meant by the term “fundamental value.” In reality, we don’t have any idea what the fundamental value of a stock is any more than we know the fundamental value of a loaf of bread. Fundamental value is just a theoretical construct. All of our models of asset prices give us a prediction of what the price should be given the relevant information. Our models are models of fundamental value. We can use our models to determine whether price is equal to fundamental value, but in doing so we are back to Fama’s joint hypothesis again.
Noise, and What do Markets Do?
The important lesson in all of this is that it helps us to understand how markets work. Markets are a discovery process. If prices fully reflected all available information, we should never observe trading other than when new information became available. Yet, we observe stocks being traded in large volumes every day. The reasons are that information is costly and people have different beliefs about what that information means.
As Fischer Black explained, it is these differing beliefs that make the market possible, but also imperfect. People use the information that they acquire to shape their beliefs about what a particular stock price should be. However, some of these traders are what Black called “noise traders.” To understand the distinction, consider the following example.
Suppose that we start with some idea of a true, underlying frictionless price for an asset. Noise traders are those who use information that they think is useful for discovering this valuation, but in reality the information they use is not relevant. Other traders in the market are informed traders in the sense that if there was some underlying frictionless price, the informed traders would use information to discover it. The trouble is that we do not know who are the informed traders and who are the noise traders!
Many have used Black’s paper to argue that financial markets are inefficient. After all, these noise traders will tend to push prices away from the frictionless price. However, Fischer Black made no such claim about inefficiency. In fact, Black emphasized that some “true” value could never be known and, more importantly, that the presence of noise traders meant that our tests of market efficiency should be much more liberal with the definition of efficiency precisely because of these noise traders.
Financial markets and the frequent trading therein are made possible by the differing beliefs among traders about what the price should be. People who have unprofitable strategies and beliefs will be pushed out of the market, while those who are successful will remain and new participants will enter. People with incorrect beliefs will also both create profit opportunities for others.
It is the competitive process within the market that drives price discovery. At any given point in time, someone might be able to beat the market. However, what the efficient markets hypothesis says is that this strategy of producing excess returns will not last forever. Either the profit is produced from exploiting some spurious relationship or the information will spread throughout the market and excess returns will be arbitraged away.
Whether this price discovery process leads to the “correct” price is not a question that can be answered. As a result, it might seem like the EMH is untestable. However, for me, the best test of the EMH is a direct test of what it really says: that people cannot systematically beat the market. To test this directly, one could compare the performance of active money managers. Surely, if anyone can systematically beat the market, it is the people who specialize in allocating money. Evidence that money managers can outperform the market in significant enough numbers to rule out randomness would seem to be a pretty strong refutation of the EMH. When you examine the performance of professional managers (see here and here), it is hard to reject the hypothesis that financial markets are efficient.