September 19, 2008

Scientists Predict the Next President

Two mathematicians have devised what they say is a "surprisingly
effective" means to predict the outcome of the U.S. presidential
election using median statistics based on voter polling.

In a paper in the journal Mathematical and Computer Modeling, Wes
Colley, of the University of Alabama, Huntsville, and J. Richard Gott
III, at Princeton University, said they have developed a system that
uses the margins of victory for each candidate in each of the many
polls taken during the past month. Those margins are then ranked from
the largest margin to the smallest, and the middle number, or the
median, is used as the candidate's score for the individual states.

Colley is nationally known for his computer modeling system used in
determining college football rankings by the NCAA, and his new system
has Barack Obama ahead of John McCain.

"John McCain needs several swing states to break his way, whereas
Obama can afford to lose a couple and still win the election," Colley

Before the Obama supporters start breaking out the champagne,
however, they should be aware of new research from the Maryland-based
Institute for Operations Research and the Management Sciences
(INFORMS), the world's largest society for operations research

Using a methodology that "applies a mathematical model of state
polling data, using a dynamic programming algorithm to forecast
electoral results," INFORMS has McCain ahead by as many as 27 electoral
votes (282.8 votes for McCain, 255.2 for Obama).

Then there is University of New Hampshire Survey Center founder and
former Gallup Poll managing editor David Moore, who is revealing in a
new book that "media polls
are not used to uncover the 'will' or thoughts of the public, but
rather to manufacture a 'public opinion' that grabs the attention of
journalists and can be used to fill media news holes." The methodology
used by the major national polls, he says, "give false readings of
which candidates voters prefer and what the public wants."

Inside Science News Service is supported by the American Institute of Physics.