Misbehaving: The Making of Behavioral Economics

by Richard H. Thaler

  The efficient market hypothesis could be reconciled with our results if the Loser stocks had high betas and thus were risky according to the CAPM, and the Winner stocks had low betas, meaning they were less risky. But we had already checked this out ourselves and reported the results in the paper; in fact, we had found the opposite pattern. For example, in the tests we ran using Winner and Loser portfolios based on three-year “formation periods” and followed by three-year “test periods,” the average beta for the Winners was 1.37 and for the Losers was 1.03. So the Winners were actually riskier than the Losers. Adjusting for risk using the standard methods of the profession made our anomalous findings even more anomalous!

  To rescue the no-free-lunch aspect of the EMH, someone would have to come up with another way to show that the Loser portfolio was riskier than the Winner portfolio. The same would be true for any measure of “value,” such as low price/earnings ratios or low ratios of the stock price to its book value of assets, an accounting measure that represents, in principle, what shareholders would get if the company were liquidated. By whatever measure one used, “value stocks” outperformed “growth stocks,” and to the consternation of EMH advocates, the value stocks were also less risky, as measured by beta.

  It was one thing for renegades like us, portfolio managers like Dreman, and dead guys like Benjamin Graham to claim that value stocks beat the market, but this fact was only declared to be officially true when the high priest of efficient markets, Eugene Fama, and his younger colleague who would become his regular collaborator, Kenneth French, published similar findings. In part provoked by our initial findings and those of Banz, who had documented the small firm effect, in 1992 Fama and French began publishing a series of papers documenting that both value stocks and the stocks of small companies did indeed earn higher returns than predicted by the CAPM. In 1996 they officially declared the CAPM to be dead, in a paper with the provocative title “The CAPM Is Wanted, Dead or Alive.”

  While Fama and French were ready to declare the CAPM dead, they were not ready to abandon market efficiency. Instead, they proposed what is now known as the Fama–French Three Factor Model, in which, in addition to the traditional beta, two extra explanatory factors were added to rationalize the anomalous high returns to small companies and value stocks. Fama and French showed that the returns on value stocks are correlated, meaning that a value stock will tend to do well when other value stocks are doing well, and that the same is true for small-cap stocks. But Fama and French were forthright in conceding that they did not have any theory to explain why size and value should be risk factors. Unlike the capital asset pricing model, which was intended to be a normative theory of asset prices based on rational behavior by investors, there was no theoretical reason to believe that size and value should predict returns. Those factors were used because empirical research had shown them to matter.

  To this day, there is no evidence that a portfolio of small firms or value firms is observably riskier than a portfolio of large growth stocks. In my mind, a paper titled “Contrarian Investment, Extrapolation, and Risk” published in 1994 by financial economists Josef Lakonishok, Andrei Shleifer, and Robert Vishny settled any remaining questions about whether value stocks are riskier. They are not. It also convinced the authors of the paper, since they later started a highly successful money management firm, LSV Asset Management, which is based on value investing.

  Although their paper convinced me, it did not convince Fama and French, and the debate has continued for years as to whether value stocks are mispriced, as behavioralists argue, or risky, as rationalists claim. The topic is still debated, and even Fama concedes that it is impossible to say whether the higher returns earned by value stocks are due to risk or overreaction. But in late-breaking news, Fama and French have announced a new five-factor model. The new factors are one that measures a firm’s profitability (which predicts high returns) and another that captures how aggressively a firm invests (which predicts low returns). In a nice twist of fate, profitability is another trait that Benjamin Graham looked for in judging the attractiveness of a firm as an investment. So in some ways, the venerable Ben Graham has been given a Fama–French seal of approval, since they also endorse value and profitability. And it is difficult to tell a plausible story in which highly profitable firms are riskier than firms losing money.

  So, in the time since Sharpe and Lintner created the CAPM in the early 1960s, we have gone from a one-factor model to a five-factor model, and many practitioners would add a sixth factor: momentum. Firms that have done well over the last six to twelve months tend to keep doing well for the next six to twelve months. Whether there are five or six factors, I believe that in a rational world, the only factor that would matter is the first one, good old beta, and beta is dead. And the others? In a world of Econs, they would all be SIFs.

  ________________

  * One note about some confusing terminology: In this chapter and the next one, when I use the term “mispricing” I mean that a stock price will predictably move in some direction, up or down, so much so that an investor could hypothetically take advantage of it for a “free lunch.” This is the first illustration of the subtle ways in which the two components of the EMH are intertwined. It is reasonable to think that stocks that are priced “too low” will eventually beat the market, but De Bondt and I had no conclusive evidence that the Losers’ prices diverged from their intrinsic value, just that they earned higher returns.

  † Just to avoid any confusion, I should mention that this “beta” has nothing to do with the beta in the beta–delta models of present bias in chapter 12. All I can say is that economists like Greek letters and beta comes early in the alphabet.

  24

  The Price Is Not Right

  Recall that the efficient market hypothesis has two components: you can’t beat the market (there is no free lunch), and prices are “right.” The work Werner and I did primarily questioned the first principle. Meanwhile, another battle was brewing about the rationality of the aggregate stock market that addressed the second principle. Robert Shiller, now a professor at Yale University, published a paper in 1981 with a striking result.

  To understand Shiller’s findings, it helps to first think about what should determine a stock’s price. Suppose a foundation decides to buy a share of stock today and hold it forever. In other words, they are never going to sell the stock—so the only money they will ever get back are the dividends they receive over time. The value of the stock should be equal to the “present value” of all the dividends the foundation will collect going forward for forever, meaning the amount of money that the flow would be worth, after appropriately adjusting for the fact that money tomorrow is worth less than money today.* But because we don’t exactly know how much a given stock will pay in dividends over time, the stock price is really just a forecast—the market’s expectation of the present value of all future dividend payments.

  An important property of rational forecasts—as a stock price is supposed to be—is that the predictions cannot vary more than the thing being forecast. Imagine you are trying to forecast the daily high temperature in Singapore. The weather doesn’t vary much in this Southeast Asian city-state. Typically the high temperature is around 90°F (32°C). On a really hot day it might reach 95°F. A “cold” day might top out at 85°F. You get the idea. Predicting 90°F every day would never be far off. If some highly intoxicated weather forecaster in Singapore was predicting 50°F one day—colder than it ever actually gets—and 110°F the next—hotter than it ever gets—he would be blatantly violating the rule that the predictions can’t vary more than the thing being forecast.

  Shiller’s striking result came from applying this principle to the stock market. He collected data on stock prices and dividends back to 1871. Then, starting in 1871, for each year he computed what he called the “ex post rational” forecast of the stream of future dividends that would accrue to someone who bought a portfolio of the stocks that existed at the ti
me. He did this by observing the actual dividends that got paid out and discounting them back to the year in question. After adjusting for the well-established trend that stock prices go up over long periods of time, Shiller found that the present value of dividends was, like the temperature in Singapore, highly stable. But stock prices, which we should interpret as attempts to forecast the present value of dividends, are highly variable. You can see the results in figure 12. The nearly flat line is the present value of dividends, while the line jumping around like the forecasts of a drunk weatherman is actual stock prices, both of which have be adjusted to remove the long-term upward trend.

  FIGURE 12

  Shiller titled his paper “Do Stock Prices Move Too Much to Be Justified by Subsequent Changes in Dividends?” To judge by figure 12, the answer was yes. Shiller’s results caused a firestorm in finance circles. Various papers were written attacking his methods and conclusions, one of which was gleefully heralded by critics as the “Shiller Killer.” (You may recall that one of those papers, written by Allan Kleidon, was presented at the Chicago conference discussed in chapter 17.)

  Academic economists still quibble about the right way to conduct Shiller’s test. But I believe that the debate was effectively settled a few years later, on Monday, October 19, 1987, and the days that surrounded it. That Monday, stock prices fell dramatically all around the world. The carnage started in Hong Kong and moved west, as markets opened in Europe and then the United States. In New York, prices fell over 20%, after having already fallen more than 5% the previous Friday. Crucial for our purposes, Monday the 19th was a day without any important news, financial or otherwise. No war started, no political leader was assassinated, and nothing else of note occurred. (For the sake of comparison, the U.S. stock market dropped 4.4% the day after the Japanese bombed Pearl Harbor.) Yet prices were falling precipitously all around the world. No one could say why. The volatility continued for the next few days. In the United States, the S&P 500 index of large company stocks rebounded a robust 5.3% on Tuesday, jumped another 9.1% on Wednesday, only to crash by 8.3% on Monday the 26th. The headline in the Wall Street Journal at the end of that month should have been, “Robert Shiller proven right: financial markets are too volatile.” In a rational world, prices only change in reaction to news, and during that week, the only news was that prices were moving crazily.

  If prices are too variable, then they are in some sense “wrong.” It is hard to argue that the price at the close of trading on Thursday, October 15, and the price at the close of trading the following Monday—which was more than 25% lower—can both be rational measures of intrinsic value, given the absence of news.

  When Shiller wrote his original paper, he did not think of it in psychological terms. He was merely reporting facts that were hard to rationalize. Not surprisingly, I read the paper through a behavioral lens, and saw him as a potential co-conspirator. When he came to give a talk at Cornell in the spring of 1982, he, Werner De Bondt, and I took a long walk around campus, and I encouraged him to think about his paper from what we would now call a behavioral perspective. I don’t know whether our conversation had anything to do with it, but two years later he wrote a paper that was a behavioral bombshell. The paper, titled “Stock Prices and Social Dynamics,” embraced the heretical idea that social phenomena might influence stock prices just as much as they do fashion trends. Hemlines go up and down without any apparent reason; might not stock prices be influenced in other similar ways that seem to be beyond the standard economist’s purview? Bob’s agenda in this paper was in some ways more radical than my own. Imagine trying to convince economists that fashion matters, when many have only recently retired their tweed sport jackets with leather patches. Years later, in a book with George Akerlof, Shiller would use Keynes’s term “animal spirits” to capture this notion of whimsical changes in consumer and investor attitudes.

  Although I have portrayed Shiller’s research as primarily relevant to the price-is-right aspect of the EMH, it is also relevant to the no-free-lunch component. To see why, it is useful to recall the findings about value investing. Value stocks, either those with very low price/earnings ratios or extreme past losers, predictably outperform the market. One can also compute a price/earnings ratio for the overall market. Does the same principle apply—that is, can you beat the market by buying stocks when they are relatively cheap and avoiding them when they are relatively expensive? My best answer to this question, which Shiller audaciously took on, is “Yes, but . . .”

  For an exercise like this, Shiller’s preferred method is to divide the market price of an index of stocks (such as the S&P 500) by a measure of earnings averaged over the past ten years. He prefers this long look-back at earnings because it smooths out the temporary fluctuations that come over the course of the business cycle. A plot of this ratio is shown in figure 13.

  FIGURE 13

  With the benefit of hindsight, it is easy to see from this chart what an investor would have liked to do. Notice that when the market diverges from its historical trends, eventually it reverts back to the mean. Stocks looked cheap in the 1970s and eventually recovered, and they looked expensive in the late 1990s and eventually crashed. So there appears to be some predictive power stemming from Shiller’s long-term price/earnings ratio. Which brings us to that “but.” The predictive power is not very precise.

  In 1996 Shiller and his collaborator John Campbell gave a briefing to the Federal Reserve Board warning that prices seemed dangerously high. This briefing led Alan Greenspan, then the Fed’s chairman, to give a speech in which he asked, in his usual oblique way, how one could know if investors had become “irrationally exuberant.” Bob later borrowed that phrase for the title of his best-selling book, which was fortuitously published in 2000 just as the market began its slide down. So was Shiller’s warning right or wrong?† Since his warning came four years before the market peaked, he was wrong for a long time before he was right! This lack of precision means that the long-term price/earnings ratio is far from a sure-fire way to make money. Anyone who took Shiller’s advice in 1996 and bet heavily on the market falling would have gone broke before he had a chance to cash in.

  The same is true in the housing market. One of Bob Shiller’s many admirable qualities is that he has long been an avid collector of data, from the historical data on stock prices back to 1871 that made his original paper feasible, to surveys of investor sentiment, to measures of home prices. The latter endeavor, done with his friend Chip Case, a real estate economist at Tufts University, created the now widely used Case–Shiller Home Price Index. Before Case and Shiller came along, indicators of home prices were not very reliable because the mix of homes sold in a given month could vary greatly, skewing the average. Case and Shiller had the clever idea to create an index based on repeat sales of the same home, thus controlling for the quality of the home and its location.

  A plot of the long-term growth in U.S. home prices since 1960 is shown in figure 14. The chart relies on data on home price sales collected by the government up to 2000, after which the Case-Shiller data become available so both data sources are used. All the prices are adjusted for inflation. The plot shows that home prices grew modestly for most of the period up until the mid-1990s, after which prices shot up. Furthermore, after a long period during which the ratio of the purchase price of a home to the cost of renting a similar home hovered around 20:1, home prices diverged sharply from this long-term benchmark. Looking at these data, Shiller warned of the dangers of a housing bubble, a warning that turned out to be right, eventually. But at the time, one could never be sure whether we were in a bubble or whether something in the economy had changed, causing much higher price-to-rental ratios to be the new normal.

  FIGURE 14

  I should stress that the imprecision of these forecasts does not mean they are useless. When prices diverge strongly from historical levels, in either direction, there is some predictive value in these signals. And the further prices diverge from historic lev
els, the more seriously the signals should be taken. Investors should be wary of pouring money into markets that are showing signs of being overheated, but also should not expect to be able to get rich by successfully timing the market. It is much easier to detect that we may be in a bubble than it is to say when it will pop, and investors who attempt to make money by timing market turns are rarely successful.

  Although our research paths have taken different courses, Bob Shiller and I did become friends and co-conspirators. In 1991, he and I started organizing a semiannual workshop on behavioral finance hosted by the National Bureau of Economic Research. Many of the landmark papers in behavioral finance have been presented there, and the conference has helped behavioral finance become a thriving and mainstream component of financial economics research.

  ________________

  * If the foundation ever sells the stock, then we would also include the price they get when they sell it, discounted back to the present. If they hold the stock long enough, this will have a negligible effect on the analysis.

  † For the record, I also thought that technology stocks were overpriced in the late 1990s. In an article written and published in 1999, I predicted that what we were currently experiencing would become known as the Great Internet Stock Bubble (Thaler, 1999b). But like Shiller, I would have written the same thing two years earlier if I had gotten around to it (remember, I was and remain a lazy man). Having made one correct prediction about the stock market, I am resolving not to make any more.