In these uncertain economic times, we’d all like a guaranteed investment. Here’s one: it pays a 24-cent dividend every four weeks for 60 weeks, 15 dividends in all. Then it disappears. Unlike a bond, this security has no redemption value. It simply provides guaranteed dividends. It involves no tricky derivatives or unknown risks. And it carries absolutely no danger of default. What would you pay for it?
Before financially sophisticated readers drag out their calculators, look up interest rates, and compute the present value of those future payments, I have a confession to make. You can’t buy this security, and it doesn’t really pay dividends every four weeks. It pays every four minutes, in a computer lab, to volunteers in economic experiments.
For more than two decades, economists have been running versions of the same experiment. They take a bunch of volunteers, usually undergraduates but sometimes businesspeople or graduate students; divide them into experimental groups of roughly a dozen; give each person money and shares to trade with; and pay dividends of 24 cents at the end of each of 15 rounds, each lasting a few minutes. (Sometimes the 24 cents is a flat amount; more often there’s an equal chance of getting 0, 8, 28, or 60 cents, which averages out to 24 cents.) All participants are given the same information, but they can’t talk to one another and they interact only through their trading screens. Then the researchers watch what happens, repeating the same experiment with different small groups to get a larger picture.
The great thing about a laboratory experiment is that you can control the environment. Wall Street securities carry uncertainties—more, lately, than many people expected—but this experimental security is a sure thing. “The fundamental value is unambiguously defined,” says the economist Charles Noussair, a professor at Tilburg University, in the Netherlands, who has run many of these experiments. “It’s the expected value of the future dividend stream at any given time”: 15 times 24 cents, or $3.60 at the end of the first round; 14 times 24 cents, or $3.36 at the end of the second; $3.12 at the end of the third; and so on down to zero. Participants don’t even have to do the math. They can see the total expected dividends on their computer screens.
Here, finally, is a security with security—no doubt about its true value, no hidden risks, no crazy ups and downs, no bubbles and panics. The trading price should stick close to the expected value.
At least that’s what economists would have thought before Vernon Smith, who won a 2002 Nobel Prize for developing experimental economics, first ran the test in the mid-1980s. But that’s not what happens. Again and again, in experiment after experiment, the trading price runs up way above fundamental value. Then, as the 15th round nears, it crashes. The problem doesn’t seem to be that participants are bored and fooling around. The difference between a good trading performance and a bad one is about $80 for a three-hour session, enough to motivate cash-strapped students to do their best. Besides, Noussair emphasizes, “you don’t just get random noise. You get bubbles and crashes.” Ninety percent of the time.
So much for security.
These lab results should give pause not only to people who believe in efficient markets, but also to those who think we can banish bubbles simply by curbing corruption and imposing more regulation. Asset markets, it seems, suffer from irrepressible effervescence. Bubbles happen, even in the most controlled conditions.
Experimental bubbles are particularly surprising because in laboratory markets that mimic the production of goods and services, prices rise and fall as economic theory predicts, reaching a neat equilibrium where supply meets demand. But like real-world purchasers of haircuts or refrigerators, buyers in those markets need to know only how much they themselves value the good. If the price is less than the value to you, you buy. If not, you don’t, and vice versa for sellers.
Financial assets, whether in the lab or the real world, are trickier to judge: Can I flip this security to a buyer who will pay more than I think it’s worth? In an experimental market, where the value of the security is clearly specified, “worth” shouldn’t vary with taste, cash needs, or risk calculations. Based on future dividends, you know for sure that the security’s current value is, say, $3.12. But—here’s the wrinkle—you don’t know that I’m as savvy as you are. Maybe I’m confused. Even if I’m not, you don’t know whether I know that you know it’s worth $3.12. Besides, as long as a clueless greater fool who might pay $3.50 is out there, we smart people may decide to pay $3.25 in the hope of making a profit. It doesn’t matter that we know the security is worth $3.12. For the price to track the fundamental value, says Noussair, “everybody has to know that everybody knows that everybody is rational.” That’s rarely the case. Rather, “if you put people in asset markets, the first thing they do is not try to figure out the fundamental value. They try to buy low and sell high.” That speculation creates a bubble.
In fact, the people who make the most money in these experiments aren’t the ones who stick to fundamentals. They’re the speculators who buy a lot at the beginning and sell midway through, taking advantage of “momentum traders” who jump in when the market is going up, don’t sell until it’s going down, and wind up with the least money at the end. (“I have a lot of relatives and friends who are momentum traders,” comments Noussair.) Bubbles start to pop when the momentum traders run out of money and can no longer push prices up.
But people do learn. By the third time the same group goes through a 15-round market, the bubble usually disappears. Everybody knows what the security is worth and realizes that everybody else knows the same thing. Or at least that’s what economists assumed was happening. But work that Noussair and his co-authors published in the December 2007 American Economic Review suggests that traders don’t reason that way.
In this version of the experiment, participants took part in the 15-round market four times in a row. Before each session, the researchers asked the traders what they thought would happen to prices. The first time, participants didn’t expect a bubble, but in later markets they did. With each successive session, however, they predicted that the bubble would peak later and reach a higher price than it actually did. Expecting the future to look like the past, they traded accordingly, selling earlier and at lower prices than in the previous session, hoping to realize a profit before the bubble burst. Those trades, of course, changed the market pattern. Prices were lower, and they peaked closer to the beginning of the session. By the fourth round, the price stuck close to the security’s fundamental value—not because traders were going for the rational price but because they were trying to avoid getting caught in a bubble.
“Prices converge toward fundamentals ahead of beliefs,” the economists conclude. Traders literally learn from experience, basing their expectations and behavior not on logical inference but on what has happened in the past. After enough rounds, markets work their way toward a stable price.
If experience eliminates bubbles in the lab, you might expect that more-experienced traders in the real world (or what experimental economists prefer to call “field markets”) would produce fewer financial crises. When asset markets run into trouble, maybe it’s because there are too many newbies: all those dot-com day traders, 20-something house flippers, and newly minted M.B.A.s. As Alan Greenspan told Congress in October, “It was the failure to properly price such risky assets that precipitated the crisis.” People didn’t know what they were doing. What markets need are more old hands.
Alas, once again the situation is not so simple. Even experienced traders can make big mistakes when conditions change. In research published in the June 2008 American Economic Review, Vernon Smith and his collaborators first ran the standard experiment, putting groups through the 15-round market twice. Then the researchers changed three conditions: they mixed up the groups, so participants weren’t trading with familiar faces; they increased the range of possible dividends, replacing four possible outcomes (0, 8, 28, or 60) averaging 24, with five (0, 1, 8, 28, 98) averaging 27; finally, they doubled the amount of cash and halved the number of shares in the market. The participants then completed a third round. These changes were based on previous research showing that more cash and bigger dividend spreads exacerbate bubbles.
Sure enough, under the new conditions, the experienced traders generated a bubble just as big as if they’d never been in the lab. It didn’t last quite as long, however, or involve as much volume. “Participants seem to be tacitly aware that there will be a crash,” the economists write, “and consequently exit from the market (sell) earlier, causing the crash to start earlier.” Even so, the price peaks far above the fundamental value. “Bubbles,” the economists conclude, “are the funny and unpredictable phenomena that happen on the way to the ‘rational’ predicted equilibrium if the environment is held constant long enough.”
For those of us who invest our money outside the lab, this research carries two implications.
First, beware of markets with too much cash chasing too few good deals. When the Federal Reserve cuts interest rates, it effectively frees up more cash to buy financial instruments. When lenders lower down-payment requirements, they do the same for the housing market. All that cash encourages investment mistakes.
Second, big changes can turn even experienced traders into ignorant novices. Those changes could be the rise of new industries like the dot-coms of the 1990s or new derivative securities created by slicing up and repackaging mortgages. I asked the Caltech economist Charles Plott, one of the pioneers of experimental economics, whether the recent financial crisis might have come from this kind of inexperience. “I think that’s a good thesis,” he said. With so many new instruments, “it could be that the inexperienced heads are not people but the organizations themselves. The organizations haven’t learned how to deal with the risk or identify the risk or understand the risk.”
Here the bubble experiments meet up with another large body of experimental research, first developed by Plott and his collaborators. This work explores how speculative markets can pool information from lots of people (“the wisdom of crowds”) and arrive at accurate predictions—for example, who’s going to win the presidency or the World Series. These markets work, Plott explains, because people with good information rush in early, leading prices to reflect what they know and setting a trajectory that others follow. “It’s a kind of cascade, a good cascade, just what should happen,” he says. But sometimes the process “can go bananas” and create a bubble, usually when good information is scarce and people follow leaders who don’t in fact know much.
That may be what happened on Wall Street, Plott suggests. “Now we have new instruments. We have ‘leaders,’ who one would ordinarily think know something, getting in there very aggressively and everybody cuing on them—as they have done in the past, and as markets should. But in this case, there might be a bubble.” And when you have a bubble, you will get a crash.