The Wisdom of Crowds

Several years ago I wrote about prediction markets like As the U.S. Presidential election cycle heats up I find I am drawn back to this special kind of bookmaking operation on the Internet. You can see a long list of Presidential election predictions on the Intrade site.

The phrase The Wisdom of Crowds is the title of a book by James Surowiecki, a staff writer for The New Yorker. In 2004 Surowiecki wrote that large groups of average individuals can predict outcomes with greater precision than smaller groups of experts. is a real-life, functioning demonstration of this claim.

First, a quick refresher. Anyone can start an account on the Intrade web site. You add a modest amount of money to your account using a credit card. Then you go to a specific event/market which predicts some outcome. The outcome is easy to verify, eventually. For example, there is a market for the outcome that President Obama is re-elected as president in 2012. Eventually that outcome will either be yes or no. As I write this on November 12, 2011 the prediction for this event is 52.1%. If I think it is likely that Obama will be re-elected I can buy a share in this event for $5.21. If I am right, and hold on to this share until the election I will receive $10.00 – a profit of $4.79.  If I am wrong, and hold on to the share I lose my $5.21. If events alter my prediction, I can either buy more shares for the positive outcome or sell my shares.

Here is a graph of this particular prediction and how the Intrade investors have evaluated the President’s chances. - Probability of President Obama being – Probability of President Obama being Re-Elected

You can click on the graph to see more information on this prediction. You can see that Intrade investors have gotten more pessimistic about the President’s chances over the last six months. An important thing to note is that anyone can play in this market. It is not a poll of political experts or those horrid talking heads we hear/see on broadcast media.

A Competitive Market

For my microeconomics students this is an example of a special form of a competitive market. There are many sellers (almost 2,000 to date) and an equal number of buyers. They all have approximately the same amount of information (no insider trading advantage in this case.) It is very easy to enter this market, and to leave it. There are few market imperfections – no monopoly, no obvious cartel.  If we assume, like Adam Smith did 240 years ago, that buyers and sellers will act in their own self-interest (making as much money as possible) then the market price will reach an equilibrium. That equilibrium price will change as new information arrives. For example, when Rick Perry forgets which federal agency he wants to close, some people may judge that Obama’s chances of re-election are slightly higher. They will bid the price up from $5.21 (52.1%) to something higher.

As an exercise consider Surowiecki’s claim that this large number of regular investors will more accurately predict the final outcome that a panel of experts.

Using Google Search to Measure Unemployment

So, this is cool. As reported in the Freakonomics blog, Google has a feature that allows users to correlate public patterns with search term usage. You supply data, such as initial unemployment claims, and Google will evaluate which search phrases best correlate with the data provided. Justin Wolfers reports,

I fed in the weekly numbers on initial unemployment claims—one of the most important weekly economic time series we have.  The search term that is most closely correlated? Crikey, it’s “filing for unemployment.”  Indeed, the correlation is an astounding 0.91.

Here’s the graph of time series on initial jobless claims and searches for “filing for unemployment”

Google Searches and Unemployment DataGoogle Searches and Unemployment Data