N ews and information hit the financial markets every hour of the day. A company might report earnings much higher than expected or announce a big acquisition. Traders and investors rush to digest the information and push the price to a level they think is consistent with what they have heard. But do they get it right? That is, do they react properly to the news they receive?
Recent evidence suggests investors make systematic errors in processing new information that may be profitably exploited by others. These findings are therefore a direct affront to the so- called efficient markets hypothesis - the idea that prices are right and there is no Tree lunch’ to be had. In 1978 Michael Jensen, a financial economist at Harvard Business School, wrote that ‘the efficient markets hypothesis is the best-established fact in all of social science’. Twenty years later, in part because of the findings I will outline, the debate over the efficiency of markets is in turmoil once again.
In 1985, Werner De Bondt of the University of Wisconsin and Richard Thaler of the University of Chicago Graduate School of Business caused a considerable stir by publishing an article presenting what they claimed was evidence that investors over-react to news. They found that stocks with very poor returns over a three-year period subsequently dramatically outperformed those with the highest returns over that three-year period. For example, a portfolio made up of the 35 biggest ‘losers’ earned a cumulative return 25 per cent higher than the portfolio of biggest ‘winners over the subsequent 36 months.
How might investor over-reaction explain these findings? Suppose that a company announces good news over the three years in question, such as earnings reports that are consistently above expectations. It is possible that investors over-react to such news and become excessively optimistic about the company’s prospects, pushing its stock price to very high levels and making the stock a likely candidate for inclusion in the ‘winners’ portfolio.
In the subsequent months, however, investors realize they were unduly optimistic about the business and the stock price will correct itself downwards. This correction may well lie at the root of the poor performance of ‘winner’ stocks. In a similar way,
Michael Jensen: ‘the efficient markets hypothesis is the best-established fact in all of social science’
'loser’ stocks may simply be stocks that investors have become excessively pessimistic about. As the misperception is corrected, these stocks earn high returns.
More recent studies have provided further evidence for this kind of over-reaction. Researchers have looked at stocks that are highly valued by the market — in the sense of having very high ratios of price to a measure of company fundamentals — and have found that such stocks may be too highly valued. For example, suppose you create a portfolio of stocks with very high ratios of price per share to earnings per share (often called ‘growth’ stocks) and a portfolio of stocks with very low values of this ratio (so- called ‘value’ stocks). Over a period of up to five years after the portfolios are formed, the ‘value’ stocks earn an average of almost eight per cent a year more than ‘growth’ stocks. For portfolios formed on other measures of price to fundamentals, the difference is even wider. Stocks with very low ratios of market value to book value, for example, earn an average of over 10 per cent a year more than stocks with very high market to book ratios.
The evidence so far points to investors over-reacting to information. More recent studies, however, have shown that under-reaction to information may be just as prevalent. Perhaps the most remarkable of these findings appeared in 1989. The late Victor Bernard and Jacob Thomas of Columbia University in the US grouped stocks based on the size of the surprise in their most recent earnings announcement, where surprise was measured, among other ways, relative to analyst expectations. They placed the stocks with the largest positive surprises in their earnings into a portfolio - the ‘good news’ portfolio, say - and those with the biggest negative surprises into a ‘bad news’ portfolio. They then tracked the returns of these two portfolios over the next six months. Their astonishing finding was that the ‘good news’ portfolio earned an average six-month return six per cent higher than its ‘bad news’ counterpart. This result is surprising because one would expect the stock price to reflect the good or bad earnings news immediately after the announcement. The evidence however, suggests otherwise. It points to investors under-reacting to information in the following way.
Suppose a company announces earnings that are substantially higher than expected. Investors see this as good news and send the stock price higher but for some reason not high enough. This mistake is only gradually corrected; over the next six months the stock price slowly drifts upwards towards the level it should have attained at the time of the announcement. An investor buying the stock immediately after the announcement would capture this upward drift and enjoy higher returns.
Since the publication of this study, researchers have uncovered evidence that investors under-react not only to earnings announcements but also to many other kinds of company information such as changes in dividend policy or news about share repurchase programs. For example, suppose a company says it is cutting its dividend. This is normally interpreted as bad news by the market and the stock price falls on the announcement. Recent research has found, however, that it does not fall enough at the time of announcement and instead continues to drift downwards for several months. Once again, this suggests that investors initially under-react to the bad news and only gradually incorporate its full import into the stock price.
Another well-known phenomenon believed to be related to such under-reaction is the ‘momentum’ effect. This refers to the fact that companies that have performed particularly well over the previous year continue to perform well over the next, and
those that have performed very badly continue to earn poor returns. One explanation for these results is that the companies performing well have announced good news but that investors have under-reacted to it. This mistake is only gradually corrected in the following months, leading to the continued upward drift in prices.
There are subtle differences between these studies that provide tantalizing clues about the way investors interpret information. Notice again that companies that have performed particularly badly over the previous three years subsequently reverse this trend and earn high returns. Companies that have performed badly over the previous year alone, however, do not reverse this trend but continue to do badly. Furthermore, it appears that while investors appear to under-react to isolated pieces of information, they over-react to a series of news which all points in the same direction, in other words is all good or all bad.
The preceding evidence is puzzling for those who believe the stock market is efficient, for it appears to suggest quite profitable investment strategies that verge on being ‘free lunches’ even after taking transaction costs into account. Someone who bought De Bondt/Thaler ‘loser’ portfolios and sold or took a short position in ‘winner portfolios’ earned handsome returns using this strategy over the past 70 years, as did someone who bought ‘good news’ portfolios and sold ‘bad news’ portfolios or bought ‘value’ stocks and sold ‘growth’ stocks. If investors are systematically under-reacting or over-reacting to information, there may be opportunities for exploiting these errors.
The first reaction of efficient-market enthusiasts to this evidence is to claim that it has nothing to do with investors making mistakes but simply reflects risk. In the same way that we are not surprised that stocks earn higher returns than bonds on average - they are, after all, riskier and should therefore pay us something extra on average to bear this risk - we should not be surprised if ‘loser’ portfolios earn more than ‘winner’ portfolios. It must simply be that the stocks we have grouped into the ‘loser’ portfolio are fundamentally riskier than those in the ‘winner’ portfolio. The fact that they do better on average reflects their higher risk. Similarly, under this interpretation, the ‘good news’ stocks derive their superior performance from their higher risk.
This is a reasonable argument at first sight but comes up short on closer inspection. Take for example De Bondt and Thaler’s ‘loser’ and ‘winner’ portfolios. The undeniable fact is that the average return on the ‘loser’ portfolio has been substantially higher in the historical data. What would we look for if we were trying to show that this superior performance is due to risk? We would hope to find that returns on ‘loser’ stocks, while higher on average, are also much more volatile and occasionally much worse than returns on ‘winner’ stocks.
Alternatively, we could calculate a more sophisticated measure of risk, the ‘beta’ of the strategy. Beta measures the extent to which the returns on the strategy move in line with movements in the overall market. Strategies with high betas are thought to be riskier because they offer poor opportunities for diversification.
Unfortunately, an advocate of the risk story would not have much success. It is true that ‘loser’ stocks are riskier than ‘winner’ stocks. Their returns are more volatile, they occasionally perform worse than winners and they have higher betas. However, their higher risk is not nearly sufficient to explain their higher returns. The conclusion is the same for all of the other strategies. ‘Good news’ stocks are a little riskier than ‘bad news’ stocks along the dimensions mentioned hut this is not enough to explain the difference in their average returns.
The fact that measuring risk in the ways described fails to explain the findings does not mean an immediate victory for those who believe that some stocks may be mispriced. For example, it is possible to argue that we have not measured risk properly. In a series of influential recent papers, Eugene Fama of the University of Chicago and Kenneth French of Yale University have argued that value’ stocks may be subject to important sources of risk not captured by simpler measures, such as volatility and beta, and that this explains why they have a higher average return than ‘growth’ stocks. Debate is raging, however, on the significance of the risks identified by Fama and French. There is still no consensus among financial economists on the right way to think about the relationship between risk and return and this remains an area of active research.
Another line of attack by proponents of efficient markets is to ask why the effects have not disappeared. If they are really due to human misperceptions, why have they not been quickly exploited and whittled away by more canny investors? Why, for example, have investors not rushed to buy ‘value’ stocks and to sell ‘growth’ stocks?
There are a number of responses to this. First, it is possible that investors were not aware of the potential opportunities until recently. It is true that ‘value’ stocks were advocated as good investments as early as the 1930s but the statistical reliability of the evidence has only recently been proven. Furthermore, it may be difficult for money managers to justify buying ‘loser’ or ‘value’ stocks. Such stocks have often had poor recent performance and appear more vulnerable to falling into financial distress. It is much easier to justify buying ‘winner’ or ‘growth’ stocks.
Beyond the controversy over the right way to measure risk, there is another interpretation of the findings, one that often goes by the name ‘data-mining’. Advocates of this view point out that countless hours have been spent by innumerable hopeful investors trying to unearth strategies that have historically been profitable. A typical exercise is to group stocks into portfolios based on some characteristic, in much the same way that I described earlier, and to compare the relative performance of these portfolios over time.
The problem with this approach is that if you try enough different ways of grouping stocks, you are almost certain, as a matter of chance, to come upon a strategy that has yielded high returns historically. However, there is no reason to believe that such a strategy will continue to work. It may simply be a spurious statistical artefact. While the data-mining criticism is important, it can be addressed by seeing whether the various findings can be replicated using other data, perhaps covering other periods or different countries. The evidence has proved robust to such cross-testing. A recent study by Eugene Fama and Kenneth French has found that ‘value’ stocks outperform ‘growth’ stocks in many markets throughout the world.
Even if the risk story and the data-mining critique have so far failed to explain the success of these investment strategies, the inefficient markets viewpoint faces a stiff challenge. Beyond merely presenting evidence that is consistent with investors under- reacting or over-reacting to information, it must show that this kind of investor behavior has firm foundations in human psychology. This challenge has been taken up by researchers in the new field of behavioral finance, which seeks to understand
whether aspects of human behavior and psychology might influence the way prices are set in financial markets.
As an example, consider how we might explain why - and when - people over-react to information. One explanation relies on a widely documented bias in the way individuals interpret information, known as the representativeness heuristic’. This is a broad phenomenon but can be interpreted at one level as saying that people see regular patterns and order even in completely random data.
A well-known example of this comes from a study of scoring patterns in the US professional basketball league. A typical question posed is: ‘Suppose we take two players of the same ability, one of whom has managed to score in the last three attempts, and the other who has failed in the last three attempts. Which of the two players is more likely to score on their next attempt?’ Most people pick the player who has just scored three times in a row. Underlying their choice is the belief that this player is enjoying a ‘hot streak’ that is likely to continue. The remarkable fact is that an analysis of actual scoring patterns reveals that players who have scored many times in a row are in fact no more likely to score on their next attempt than players who have repeatedly failed. That is, people see patterns and trends such as ‘hot streaks’ where none exist.
A similar argument has been used in finance to explain why people may over-react to a sequence of positive earnings announcements. In reality, changes in company earnings follow a fairly random pattern. However, when people see a company’s earnings go up several quarters in a row, they forget this means very little for the next quarter’s earnings. They wrongly believe they have spotted a trend and extrapolate recent good performance too far out into the future. Such excessive optimism pushes prices too high and produces effects consistent with over-reaction.
There are also well-known biases in human information processing that would predict under-reaction to new pieces of information. Known variously as ‘conservatism’ or ‘over-confidence’, these biases document the fact that people cling too strongly to previously held beliefs and are slow to update their views in line with new information. Clearly this corresponds directly to under-reaction to news. Investors may have views about the earnings prospects of a company and may be reluctant to abandon these upon hearing news of surprisingly high earnings. While they do push up the stock price a little, they remain sceptical about the new information and only gradually update their views to the correct extent.
While such links between psychology and finance sound plausible to many, a substantial proportion of the academic finance community views them with considerable scepticism. They accuse behavioral finance theorists of going on a ‘fishing’ expedition, sifting through texts on human psychology until they find something that could be related to the effects they are trying to explain. The most convincing models to date are, therefore, those that rely on as small a number of psychological biases as possible.
The debate over the correct interpretation of the findings highlighted here is far from being resolved. Researchers in behavioral finance are trying to build more robust and convincing models of the interplay of psychology and finance. On another front, firm believers in efficient markets are trying to understand the relationship between risk and return better in the hope this might shed light on the evidence. Nearly 20 years after Michael Jensen’s famous comment, the efficient markets hypothesis is far from being the best established fact in the social sciences.
Do investors react correctly to news? Or do they make systematic errors which can be profitably exploited by others? Module 6 gives a broad introduction to the controversial topic of behavioral finance, while in this article Nicholas Barberis explains how a whole batch of recent research has challenged the efficient markets hypothesis. He points to studies showing that investors both over-react and underreact to information, and that so-called ‘value’ stocks outperform ‘growth’ stocks. Efficient market theorists have fought back - suggesting, for example, that ‘value’ stocks may be subject to more risk than we realize and asking why the effects have not been whittled away by more canny investors. Links between psychology and finance sound plausible to many but a substantial proportion of the academic finance community views them with considerable scepticism. The debate is far from over.