The Long Tail Offline

Chris Anderson has been posting about the Long Tail, which is about how only a relatively small number of movies, records, books, etc. get wide distribution, and if you plot all of a given type of product, such as movies, on a graph with x being number of copies sold or viewed for pay, and y being individual movies in descending order of copies, you see a short fat head that gets regular distribution and a very long tail that does not. The main point is that the long tail has as much value as the fat head, and with the Internet it is now possible to distribute the long tail to cult movie viewers, old movie buffs, afficionados of certain genres, etc.

Here's a different case where everyone takes the long tail to be representative when the short head accounts for most of the money:

The first of those people was Murray Barr, and Johns and O’Bryan realized that if you totted up all his hospital bills for the ten years that he had been on the streets—as well as substance-abuse-treatment costs, doctors’ fees, and other expenses—Murray Barr probably ran up a medical bill as large as anyone in the state of Nevada.

“It cost us one million dollars not to do something about Murray,” O’Bryan said.

MILLION-DOLLAR MURRAY Why problems like homelessness may be easier to solve than to manage. by MALCOLM GLADWELL New Yorker, Issue of 2006-02-13 and 20, Posted 2006-02-06

The money in this case is medical expenses paid for by hospitals or the state, for care of homeless people.

Many people assume that everything is organized in bell curves, with most expense, population, etc. in a big hump in the middle, tapering off on each side (by age, height, or whatever). This kind of distribution is so common that it is known as the normal distribution.

But many real world distributions just don't work that way. Populations that organize themselves in power laws don't: they have a short head and a long tail. This kind of distribution has gotten most notice for networks of various types, social, biological, and technological. Mistaking a power law distribution for a normal distribution leads to bad science, bad policy, and bad economics. Maybe as people begin to see more power law distributions in non-networked policy arenas they will become less likely to mistake them in networks.

Picking the wrong distribution to describe a phenomenon is bad risk management, because it leads to bad predictions. Picking the right descriptive distribution turns out to be as important for homeless people as it is for networks.