Informational cascades

Skógafoss, Iceland

Sometimes we make our own decisions; sometimes we follow a decision taken by someone else. There can be good reasons for following another person, but if an agent makes a decision based on what another person does without regard to his own private information, just because s/he thinks that what the other decides must be right, and if other agents do so as well, then we are in an informational cascade (or information cascade).
Let’s take an example (from Noise by Daniel Kahneman et al. pp. 100-103). Ten people must decide whom to hire for an important position. There are three candidates and the members of the group announce their views in sequence. Arthur speaks first and sees Thomas as the best choice. Then Barbara speaks. She is not sure that Thomas is the best choice but is unsure who is. She trusts Arthur and so she chooses Thomas as well. Next it’s Charles’s turn. He thinks that Thomas is not a good candidate but that Julie is. Nevertheless, he also chooses Thomas, not because he feels under pressure and doesn’t have the courage to give a different opinion, but because he thinks that Arthur and Barbara have good reasons for their choices. Unless David, person 4, has strong reasons against the choice of Thomas, probably he’ll follow the others before him. If so, then we are in a cascade. Etc.
Informational cascades are also a much happening economic phenomenon. Here is an example, taken from the Investopedia: Someone might think that a certain financial pundit has more knowledge and information than he himself has. Therefore, he imitates the pundit’s stock picks. A neighbour observes him boasting about his stock picks, and she also picks the same stocks. Another neighbour notices that both people chose the same stocks and assumes that those stocks must be good picks, simply because more than one person has picked them. An informational cascade has begun, and all of the participants have very little information to back up their decision-making.
Informational cascades also happen in daily life. A group of friends wants to go out to eat together. Each one of the friends has his or her preferences where to go. One proposes a restaurant and the others agree.
There are several conditions that make that we can speak of an informational cascade (see here and here):
- There is a decision to be made; agents make the decision sequentially, and each agent can observe the choices made by those who acted earlier.
- A limited action space exists (e.g. an adopt/reject decision).
- Agents make decisions rationally based on the – usually limited – information they have.
- Agents do not have access to the private information of others.
Note that each decision is based on the information the agent has and that the agent is not deciding under group pressure. Maybe s/he has some doubts about the group decision, but in fact s/he trusts that the others are right and actually the agent lacks good reasons that they are not. Otherwise s/he would have said so. From this point of view, it’s not necessarily irrational for the later deciding agents to follow the decision others have already expressed. If you are not really sure what to do, it may be smart to follow the others.
Nevertheless, an informational cascade is not unproblematic. First, if another agent would have expressed his or her choice first, another decision might have been taken by the group; or, second, if the first agent also was not really sure what to decide, s/he might have taken another decision. So the final decision can be the result of the arbitrary decision of the person speaking first. Third, people tend to ignore that other agents are in the same informational cascade and are not deciding independently. They may think that there is some crowd wisdom, while in fact they follow the initial view of the first decider. Fourth, the decision of the first decider may be completely wrong with possible nasty consequences if everybody follows this person.
Although the decision procedure just sketched is far from uncommon, it is risky. Note that it is a procedure that is applied in a situation in which the agents are not really sure what to do and that their decisions are sequentially taken and known to everybody. In such a situation the final decision tends to be arbitrary and can even have fatal consequences. Market leaders can be the source of much financial if not economic harm, if others mistakenly think that they have the right information and therefore follow them. It is also possible that a decider is not reliable, and a non-reliable agent who wants to manipulate a decision to his or her advantage can try to be the one who decides first, hoping that others will follow. Or, a committee can take the wrong decision, such as hiring the wrong person, if it uses a sequential decision procedure and gets into a cascade. Or friends can go to the wrong restaurant. One solution to avoid an informational cascade is to vote by secret ballot, but this doesn’t work in a financial market. Then we come to such solutions as think twice before you act. However, informational cascades can also have positive effects. The fall of the German Democratic Republic in 1989 was the result of initially small protests in Leipzig that gradually grew because more and more people decided to join, just because others did. People followed other people without really knowing whether to do so was right, in the sense that the protests could have led to a massacre as well. Informational cascades are fragile phenomena.

Sources: The references quoted in the text.