Idea Epidemics and Information Cascades
Variations of the SIR model can generate chaos. Chaos theory in mathematics
shows that many nonlinear differential equation models can be chaotic in a
precise mathematical sense. That is, the system can generate seemingly random
variations—variations that never repeat themselves, that appear to be generating
random numbers even though the system is deterministic. In fact, random
number generators on computers are not really invoking chance but are the
product of such chaotic deterministic models. Variations of the SEIR epidemic
model can be chaotic, as has been shown and studied mathematically and related
to actual disease data.21
Chaos theory is associated with the butterfly effect, which refers to the idea
that a huge, apparently unpredictable storm might have been generated by a
seemingly distant and irrelevant event such as a butterfly flapping its wings on
the other side of the planet long ago. Another variation of the SIR model can
help explain such butterfly effects by adding information cascades to the basic
model.22 If people think they are collecting reliable information by observing the
numbers of people who make certain choices, then the equilibrium can move off
in random directions, much as in the artificial music-market experiment of
Salganik and his colleagues discussed in chapter 4. I recall an experience with
Professor Ivo Welch of UCLA, one of the authors of the information cascade
theory. While driving me to my hotel, he told me he thought we were near the
hotel but that he wasn’t sure exactly where it was. Then he spotted a taxi with no
passenger, and he said that he would just follow the taxi, because there was a
good chance that the taxi was on its way to the hotel. His guess that the taxi
driver had the information we needed worked perfectly, but it could just as well
have led us to a different hotel or to any number of random places. If a lot of
people were behaving as Ivo was, then one initial taxi could, in principle, start an
epidemic that could set off a deluge of taxis to a random place.
Information cascades can explain how speculative bubbles can be perfectly
rational, in accordance with the canon of economic theory. In my view, they are
interesting because they describe how bubbles or depressions can start from
purely random causes, even if people are fairly sensible. George A. Akerlof and
Janet L. Yellen coined the term “near-rational” in 1985, and I wish that term had
caught on more, that it had gone viral.23 However, information cascades may not

be so important a problem. In reality, taxi drivers never seem to follow the
leader, at least not in terms of driving to destinations in their city. But, like
everyone else, taxi drivers may follow others in terms of remembering “facts” of
a more ambiguous nature, such as the best restaurant in a city.24 Ask a taxi driver
to take you to the best restaurant: you will likely get laughter in response, and it
is unlikely that the destination will be demonstrably the best.25
The movements of taxi drivers, just like changes in behavior of consumers,
investors, and entrepreneurs and other economic phenomena, can never be
properly understood without some input from narrative economics. Making real
progress in narrative economics is a big project for serious research in the future.

Notes