Informational Cascades and the Financial Crisis

Was the financial crisis caused by an informational cascade?:

Wall Street's Lemmings, by Cass R. Sunstein, TNR: ...It is impossible to understand the ... current financial crisis ... without exploring the dynamics of informational cascades. ...

To get a sense of how cascades work, imagine that a group of people is deciding whether to invest in real estate or instead the stock market. Assume that group members are announcing their views in sequence. From his own knowledge and experience, each member has some private information about what should be done. But each member also attends, reasonably enough, to the judgments of others.

Andrews is the first to speak. He suggests that real estate is the way to go. Barnes now knows Andrews's judgment;... if her independent judgment is otherwise, she would--if she trusts Andrews no more and no less than she trusts herself--be indifferent..., and she might simply flip a coin.

Now turn to a third person, Carlton. Suppose that both Andrews and Barnes have favored investing real estate, but that Carlton's own information, though not conclusive, suggests that this is a definite mistake. Even in that event, Carlton might ... follow Andrews and Barnes. ... If he does, Carlton is in a cascade.

Now suppose that Davis, Eagleton, and Franklin know what Andrews, Barnes, and Carlton said. On reasonable assumptions, they will do exactly what Carlton did ... regardless of their private information... And all this will happen even if Andrews initially blundered. ... That initial blunder can start a process by which a number of people end up making really terrible decisions. Human beings ... can lead one another to disaster. ...

I have given a highly stylized example, but informational cascades are pervasive in the real world. Why do restaurants suddenly become immensely popular, when equally good restaurants are collapsing? Why do some songs, books, and movies become huge hits...? ... A large part of the answer involves the power of some initial sparks, which initiate a cascade. ...

Turn in this light to the current crisis. By historical standards, home prices jumped spectacularly from 1997 to 2004. In that period, many people thought ... that it is in the nature of home prices to increase over time, and people's behavior tracked their belief. But the belief was demonstrably false. From 1960 to 1997, home prices were relatively stable, until the unprecedented boom that began in 1997.

As behavioral economist Robert Shiller has shown, ... people were greatly influenced by a process of social contagion that amounted to an informational cascade. ...

In 2005, Shiller and Karl Case conducted a survey among San Francisco home buyers. The median expected price increase, over the next decade, was nine percent per year! ... Their baseless optimism was based on two factors: salient price increases in the recent past and the apparent, and contagious, optimism of other people.

Of course the stock of public knowledge depends ... on the media as well. ... If the apparent experts confirm "what everyone knows," then seemingly risky deals, of the sort that have led so many people to disaster, will seem hard to resist. ...

It is true that an understanding of informational cascades does not by itself compel any specific set of policy prescriptions. But it does suggest that policymakers need to have a sense of not only of economics but also of public psychology and of the power of social interactions--and their potential role in shaping them. The nation's greatest leader during a time of economic distress, Franklin Delano Roosevelt, famously said during his first inaugural address, "The only thing we have to fear is fear itself." Throughout the Great Depression, Roosevelt showed a keen intuitive understanding, not of economic theory, but of the need to inculcate a sense of public confidence and to stem public panic. His own success stemmed largely from his ability to assure the nation that it was in the hands of a leader who was at once competent, organized, confident, and calm. The Bush administration does not appear capable of following his example. The next administration had better try.

Comments

[Techie]

I suppose that trying to stop this slander on lemmings is hopeless. However, they don't do that over the cliff thing in the real world; that's something Disney made up out of whole cloth back in the 1940s.

I suggest that in the future we use Wall Streeters as our example of suicidal crowd behavior instead.

[baileyman]

But this is not really talking about "information", is it? I mean, none of this constitutes actionable news, right?

Although I will say that I wrote my congressman to support the revised Paulson bill because Buffett said so (thereby making it "The Buffett Bill"), noting that there was still the possibility we were being railroaded and that the whole thing stunk.

[Noni Mausa]

Thanks, Techie. We usually think of false animal metaphors as being mediaeval, but we have plenty of our own.

I participated in a classroom project modeling the allocation of funds to community projects by a "city council" of students. There were five or six groups of us, and at the end of the allocation process we all presented our results.

In most groups, the members discussed how the money should be allocated among several choices -- senior community support, teen job training, a battered women's shelter, a swimming pool, a school breakfast program, cleanup of a local lake, that sort of stuff. In these groups, the final allocation of the money was very consistent from group to group.

I suggested a different approach in our group, similar to a silent auction. Each "councilor" was given control over a percentage of the available funds, and decided how much of his or her money should go to which project. Then we pooled our results to reach a final result that was radically different from the other groups' decisions.

A third approach I had considered (but our group agreed would take too long) was to view the projects not as competing, but as potentially a cascade where injection of funds at one point (perhaps teen jobs) could cascade down into the others (senior home support, lake cleanup, school lunch preparation) thus getting several uses out of each dollar.

My point? To a large extent the process used for deciding something determines what decisions can be made. Different process -- different decisions. Not an earthshaking concept, but it keeps getting forgot.

Noni

[robertdfeinman]

This is a restatement of what George Soros calls "reflexivity". The information that is believed by crowds feeds back and changes the future reality.

On Bill Moyer's latest show he was asked about how long the crisis will last and he replied, in effect, it depends upon what actions are taken now.

The idea that humans can influence the future keeps being rediscovered as if it were some great new insight...

[bakho]

In an environment with too much information, people will access the distillation of the information, not the raw information itself. If the distillation is missing a key piece of information, it is more difficult to know what is missing.

In the above scenario, participants are only comparing their conclusions. They are not comparing information, assumptions and inferences. The speakers will more likely cascade bad conclusions if those conclusions are based on large complex sets of information.

To go along with a conclusion that one does not understand and without evaluating the basis of the conclusion violates intellectual courage, intellectual autonomy, integrity, perseverance, confidence in reason and fairmindedness.

[Roger Chittum]

Good point, Bakho. It readily becomes the wisdom of mobs.

[Mike]

The problem with cascades is exactly that people take actions that fundamental information does not rationally support. It isn't a problem of the quantity of information. I do think that cascades can be perpetuated or triggered more easily when decision makers have a difficult time telling the difference between fundamental information and someone else's interpretation of fundamental information. We live in a society where news companies have an incentive to sell their news, and it turns out that sensational headlines sell more news than unsensational headlines, and so news companies often fail to report unadulterated fundamental information, but rather interpretations that tend to sell more news. In times of financial crisis, that's bad.

[Ninja Zombie]

Indeed, it is hard to see the public policy implications of information cascades. Policy makers often fall victim to the exact same cascade.

One of my favorite examples of an information cascade is the Copenhagen interpretation of quantum mechanics. Copenhagen falls into the category of "not even wrong": no one can precisely define a measurement beyond "I know it when I see it", but measurements are a fundamental part of Copenhagen QM.

Every physicist studying quantum measurements knows this, as do the quantum information people. Nevertheless, the "consensus" among physicists is that Copenhagen is correct (go pick up any book titled "Introduction to Quantum Mechanics"). This situation came about because the Copenhagen school had lots of grad students.

The reaction of policy makers (the NSF and also PRL*) is not helpful. They reinforce the cascade. Anyone who doesn't believe in Copenhagen is a crackpot and should not be given a grant.

Just an example to consider.

(*) PRL is a privately run journal. However, due to it's central role in the physics community, it plays a role much like a policy maker.

[Bruce Wilder]

Mostly, I suspect that successful restaurants are good. Information cascades might amplify their success, but not their "goodness", and that could be a useful insight.

But, Sunstein is making a much stronger assertion, and one that is ungrounded: that there is no justice in the world, and information cascades are a reason.

He argues that there are good restaurants that fail, which is undoubtedly true. Many a good product has failed, for a failure of marketing. It is a seductive argument, because we know that God's rain falls on the fields of the apostate equally with the fields of the saint.

But, do we really want to believe that lemmings are suicidal? Why is that an appealing myth?

Sunstein is proposing not a theory, but a myth, and we would do well to turn away.

[Alex Tolley]

The informational cascade mechanism sounds very Bayesian to me, and there is evidence that this is how the brain works - efficiently able to make fast decisions which have evolutionary value in a dangerous environment.

The other point is that everyone in the chain observes the results and this makes it hard to resist the conclusion because the individual looks foolish missing the rewards that the group is acquiring in the short term.

[Bruce Wilder]

AT: "the individual looks foolish missing the rewards that the group is acquiring in the short term."

Yeah, that Krugman sure is shrill.

[Julio]

Does this completely ignore the issue of feedback?

I.e., restaurants don't get better because more people like them (even this is arguable, your experience of the restaurant might actually be better if it's popular). But real estate does go up if people think it will.

"Their baseless optimism was based on two factors: salient price increases in the recent past and the apparent, and contagious, optimism of other people."

So, seeing prices rise, more and more people putting more and more money into the market, and thinking prices will rise some more is "baseless optimism"?

People expected prices to rise 9%/yr. for 10 years; was that wrong and foolish ("baseless") very early in the bubble?

The kind of information cascade the author depicts is an interesting field of study, but doesn't seem too applicable here.

[KThomas]

"People expected prices to rise 9%/yr. for 10 years; was that wrong and foolish ("baseless") very early in the bubble?"

Yes, Julio. I had people telling me it would go up 20%, per year. Family members, friends, they all told me to buy now, or suffer. I ignored them. Nothing goes up 20%, per year. That's called Bull Shit.

[Alex Tolley]

kthomas: "Yes, Julio. I had people telling me it would go up 20%, per year. Family members, friends, they all told me to buy now, or suffer. I ignored them. Nothing goes up 20%, per year. That's called Bull Shit."

But if houses go up 20% a year for 2 years before the price rises stall, that makes a house substantially more costly in 2 years to a buyer with say 3%/yr salary raises. Jumping in as the realtors exhorted could make sense in that scenario,?

market price trends also go on for longer than seem rational, but betting against them, by going short can easily wipe you out - just ask the Feshbach brothers who were huge shorts in the early 1990's.