Dan Moran: Aggregators cause poll aggravation
November 5, 2012 6:28PM
Updated: December 7, 2012 6:15AM
In Gore Vidal’s “Lincoln,” we’re told that when the Railsplitter learned he lost not only his native state of Kentucky, but also his home county of Sangamon in the 1864 election, he was ready with a joke: “It would appear where I’m best known, I’m least popular.”
Honest Abe didn’t seem surprised by those results, even though he was speaking 114 years before the birth of Nate Silver.
Who is Nate Silver, you ask? He is one of the new-age titans of “poll aggregators” — the stats’ geeks who run raw numbers through computer programs to forecast election results. Silver’s legend began to grow four years ago this November, when his Web site/blog, 538.com, correctly predicted 49 of the 50 state results in the presidential race (his one miss: Indiana) and all of the Senate races.
In 2012, as of Monday afternoon, 538.com gives Barack Obama an 86.3 percent of winning at least the required 270 votes in the Electoral College, with a forecast of 307.2 electoral votes and 50.6 percent of the popular vote.
But Silver has critics — mostly, this year, on the right — who point out that his formula is not infallible. Locally, the evidence includes U.S. Rep. Joe Walsh. On Nov. 2, 2010, 538.com gave incumbent Democrat Melissa Bean an 88.2 percent chance of winning in the 8th Congressional District, with a projected election result (based on “polling, expert forecasts, fund-raising, past election returns and other indicators”) of 52.9 percent of the vote for Bean and 45.5 percent for Walsh.
The result, was, of course, much different — 48.5 percent for Walsh, 48.3 percent for Bean and, lest we forget, 3.2 percent for the Green Party’s Bill Scheurer.
This being the age of smartphones that are more powerful than the guidance computer on Apollo 11, 538.com isn’t the only poll aggregator in the game, and it isn’t the only one forecasting an Obama win in 2012. The Princeton Election Consortium, a “meta-analysis” of state polls run by neuroscientist Sam Wang, correctly predicted George W. Bush’s 286 electoral votes in 2004 and has Obama finishing with 303 electoral votes today.
Other aggregator sites predicting an Obama victory include Pollster.com (at least 270 electoral votes) and Votamatic.org (326). Meanwhile, Real Clear Politics assigned 201 electoral votes to Obama and 191 to Mitt Romney on Monday, listing such states as Michigan and Pennsylvania in its toss-up category.
Fortunately for those of us whose brains explode when it comes to probabilities and statistics, we can wait for someone else to do the most understandable and final math tonight.