When two seemingly similar restaurants sit across the street from each other, one packed with customers and the other empty, why do so many diners unfamiliar with either assume the crowded restaurant is the better one?
People often chalk up acting on such assumptions to a herd mentality, a mindless tendency to follow the crowd or to conform. But herd behavior is not as mindless as it seems to be, explains Professor Professor Boğaçhan Çelen, whose research often examines uniform social behavior.
In an information-based explanation of herd behavior, an individual weighs his or her own private information, such as past experience or research, as well as that of others, such as recommendations, before following the crowd.
A prospective diner who knows little about either of two restaurants may at first consider choosing the empty restaurant, A. But after noticing the popularity of the other restaurant, B, he or she may conclude that customers at the more popular restaurant must have superior information and that it must therefore be better. The crowded restaurant is not necessarily the superior establishment, because early diners have chosen it based on “wrong” information; Even if as little as 10 percent of all diners believe B superior, if enough such diners happen to show up in succession early in the evening, subsequent diners with good information about A and no previous knowledge of B may observe the growing crowd at B and assume diners there have better information.
In cascade behavior, an individual wholly disregards his or her own information to follow the crowd. Diners deciding between two restaurants may have done some research suggesting that the empty restaurant is the better of the two, but still choose the crowded restaurant even though their private information indicates the superiority of the empty restaurant. This is simply because the strength of the information that is revealed by others dominates their own private information.
“The information collected from the actions of others overwhelms and suppresses one’s own information,” Celen explains. “Often, that new information isn’t good information.” Another way to think about the difference between herds and cascades is that when acting in a herd individuals may have decided differently with new or different private information. In a cascade, the belief that a group must be acting optimally overrides all else, including new or contradictory private information.
Previous research examining herd behavior did not recognize a distinction between herds and cascades; Çelen and coresearcher Shachar Kariv of University of California, Berkeley, set out to demonstrate that distinction and learn more about the thought process that drives decision making and how people use information to form beliefs.
Given that individuals’ beliefs can’t be observed, analyzing the role of information in decision making is a difficult task. Çelen and Kariv set up an experiment that allowed them to distinguish between herd and cascade behavior and elicit the beliefs of subjects.
Subjects were asked to make mock investment decisions over several continuous rounds in which they could observe others’ decisions. The researchers observed whether subjects based their decisions solely on their own information or on that of other subjects, and whether subjects would change their decisions after learning about the actions of other subjects.
“The information people have gives them some idea of how they should proceed, but it’s not necessarily accurate,” Çelen says, “So when they observe other people going in a different direction than what they were going to do, they may reevaluate their view.”
In more than one-third of rounds the subjects exhibited herd behavior, in which all individual subjects made investment decisions suggesting that they believed the information they garnered from the other subjects was better than their own private information. Another one-third of the time the subjects exhibited cascade behavior, in which each subject, despite conflicting private information, made the same decision as the previous subject.
“Understanding cascades is important,” Çelen explains, “because when we’re headed towards a bubble, an authority with more valuable information, such as the government, can step in and publicize that information, encouraging people to defuse a bubble before it forms.”
The dot-com boom of the 1990s is a prime example of the dangers of cascade behavior. Internet ventures multiplied, and although available information pointed to a looming bust, investors and dot-com entrepreneurs ignored those warning signs and forged ahead before the bubble burst. Likewise, the subprime crisis played a key role in the housing market bubble, and in turn fed a second bubble as large financial institutions continued to make high-risk investment decisions backed by bad loans in the face of contrary signs suggesting that those investments were unsound.
Herd behavior always retains the possibility that individuals will act on their own information, which clues in others to potentially new information on which to base future actions. In cascades, where no individuals give credence to their own information, no such new information can surface.
“When people start ignoring their own information,” Çelen says, “no new information is transmitted to financial markets, creating a bubble effect. Getting information from a variety of sources is always valuable because it creates a healthier, more diverse market.”Boğaçhan Çelen is associate professor of finance and economics at Columbia Business School.
Read the Research
"Distinguishing Informational Cascades from Herd Behavior in the Laboratory"