An informational cascade occurs when imperfectly informed people pay attention to the signals given by the views of earlier speakers or actors; those imperfectly informed people join the cascade, thus amplifying the very signals by which they were influenced. If Adam thinks that global warming is a serious problem, Barbara may go along, and Charles may be reluctant to reject the shared view of Adam and Barbara; once Adam, Barbara, and Charles say, and think, that global warming is a serious problem, and once Danielle goes along too, a cascade is starting.
A reputational cascade occurs when people silence themselves because they don't want to be criticized, punished, or ostracized. Those who silence themselves, or seem to go along with the crowd, do not weaken and may even strengthen the reputational pressure on others. If Adam thinks that global warming is a serious problem, Barbara may go along for fear of incurring Adam's disapproval, and Charles may be reluctant to take on both Adam and Barbara; once Danielle goes along too, a reputational cascade is starting.
We are witnessing a Miers cascade, in which many people are giving negative or lukewarm signals largely because of (a) the information contained in the previous signals of others or (b) the reputational pressure to give negative or lukewarm signals. I don't mean to make any evaluation of Ms. Miers' nomination, or even to predict the outcome. But it is clear that we are in the midst of a textbook example of a cascade.
I'm not sure this observation comes to much without more. While participating in a cascade may be epistemologically irresponsible, whether such participation has undesirable effects depends upon whether the views expressed or withheld by Barbara and Charles are true. As long as one remains agnostic about the Miers' nomination, as you do, one cannot determine whether the textbook cascade you describe is positive or negative.
Posted by: Troy | October 26, 2005 at 11:17 AM
I think these cascades are presumptively negative. Many issues that are important are decided by a consensus of opinion, rather than by accurate study.
For instance, most people think DDT is a cancer causing chemical that drives birds to extinction.
These are both scientifically questionable claims, but it is tough to fight the "Oh, come on" reaction to any claim that goes against the public consensus. So when you have an uninformed cascade, it has the effect of making unsettled questions appear to be settled. This raises the burden of proof very high for some claims, when under an objective standard the claim is at least plausible.
Posted by: D Conrad | October 26, 2005 at 11:55 AM
Professor Sunstein,
If the urn contains 100 red balls, and 0 white, is it really a cascade when A, B, and C say the urn is red dominated?
Posted by: Rue Des Quatre Vents | October 26, 2005 at 01:54 PM
I agree opinion cascades are fascinating but I'm not sure we have one here. Many people are giving lukewarm or negative signals, but how do we know it's because of reasons (a) or (b)? "Cascades" in opinions seem most likely to occur when the value of holding similar opinions is relatively high but the intrinsic value of holding those opinions is relatively low. There's some evidence that, at least among Republicans, the opposite has occurred: the attraction or standardization value is being outweighed by the desire for a substantively beneficial result. The expression of lukewarm or mildly negative views may be part of the process of forming of a mixed equilibrium where individuals settle on different views. Or there might be a slow shift toward one pole as people are persuaded, on the merits of information being dispersed and with little influence from positive feedback.
Posted by: Bruce | October 26, 2005 at 02:43 PM
I second the previous two commenters. What does it mean to classify a thing as a "cascade?" Either "cascade" implies something about the validity of the views that are cascading (i.e. they are less likely to be rational) or it doesn't. If it doesn't, then classifying something as a "cascade" is wholly uninteresting. If it does, then the classification of something as a "cascade" requires an evaluation of whether the people with cascading views are nonetheless making use of rational decisionmaking processes or not.
Either way, I don't know what saying "look, this is a cascade" with nothing more adds to the discussion.
Posted by: Paul Gowder | October 26, 2005 at 03:22 PM
Also, how do we know that "people [have been" silenc[ing] themselves because they don't want to be criticized, punished, or ostracized?" This statment, the apparent predicate for a "reputational cascade" assumes both that (A) people hold the opinion that Miers is qualified, and (B) the people described in A aren't talking because of pressure. The problem is that absent careful and confidential polling or some similar method, the presence of (B) precludes finding evidence for (A). We have no way to know whether the Miers-supporters are silent because they're cowed or because they don't exist.
Posted by: Paul Gowder | October 26, 2005 at 03:26 PM
Yeeeeaahhh SOUTH SIDE!!!!
WHITE SOX RULE!!!!!!!!
Posted by: Steve Bartman | October 26, 2005 at 11:10 PM
Actually, the notable part is that there is much more GOP opposition to Miers than the press is reporting. The White House had lined up an army of commentators who were supposed to push whoever Bush nominated. When they found out it was Miers, they mostly agreed to keep quiet: no one would speak in her favor, but no one wanted to tick off the White House by saying what they actually felt -- namely, that they don't want her on the court. Here's a test: Try coming to DC, going to a GOP hangout like Capital Grille, and just say the word "Miers." People will flip out; it's the "M word" of GOP circles.
Posted by: DC GOP lawyer | October 26, 2005 at 11:20 PM
What is the relationship between information cascades and the Condorcet Jury Theorem? As I understand it (from this post by Prof. Sunstein http://www.lessig.org/blog/archives/003027.shtml), that describes the situation where each individual member of the crowd has a less than 50% chance of correctly forecasting the event, then the collective judgment will almost certainly be wrong. Isn't an accurate description of an information cascade one where the majority of participants do not have an independent ability to predict the outcome, and are further largely relying on others with the same deficiency?
Posted by: jleons17 | October 28, 2005 at 01:58 PM