Data Projected Obama Winner Well Before Final Count

By now, the final count between Barack Obama and John McCain is well known. Yesterday, however, it remained a tricky game of guesswork among the news media and political pundits.

On the web, tech giants and betting exchanges strained lucid answers out of their own user data.

Facebook's Lexicon tool - which analyzes the frequency and associations of words in Facebook wall messages - showed that Obama was mentioned significantly more frequently than Mccain. More importantly, mentions for Obama were more positive (e.g. "I love Obama"), versus negative (e.g. "I hate McCain"). On the graph below, a 70% score means that 70% of the mentions had a positive sentiment and 30% had a negative sentiment.

It also appeared that Yahoo was making an official call on the election via its Political Dashboard service, which unlike Google's Map Mashup and CNN's ElectionCenter 2008 went beyond real time tracking and projected Obama to win, but by a slightly narrower margin than was the reality:

The dashboard, however, is not powered by Yahoo directly but by Real Clear Politics, who reported Obama far ahead in their own polls, that closed Nov. 3 - as clarified by Search Engine Journal. Yahoo did have a Search Tally that showed Barack Obama topping John McCain in searches, 64.8% to 35.2%.

Probably the most interesting projection was from, a costume and mask site. They tracked the sales of Obama and McCain $0.99 paper masks (1 mask = 1 vote) and declared Obama-Biden the winner. The Presidential Mask Poll claims to have accurately predicted the winning ticket since 2000 - even when most of the media got it wrong. Though their 55% to 45% result was nowhere near the electoral landslide reality, it was much closer to the popular vote - 52% to 46%.

Predictions from the election betting sites were not so modest: Dublin-based InTrade and Betfair in the UK - the same sites that gave George W. Bush 2-to-1 odds of beating Sen. John Kerry in 2004 - gave Obama odds of winning at 92% and 94% and McCain just 9% and 7%, respectively.

Betting exchanges tend to be surprisingly accurate in their predictions, writes CNET. When there is a financial penalty for making a mistake, results tend to be a bit more accurate than when the only punishment for being wrong is potential embarrassment. Two economics professors working on a paper about the international history of betting markets (pdf) once calculated that between 1884 and 1940, betting markets predicted the popular-vote winner in every presidential race but one.

Often, however, they are better at "reflecting conventional wisdom" than predicting the future - Intrade called New Hampshire for Obama (54%) right before Hillary Clinton took it, and in the winter of '07, it had Rudy Giuliani's chances of winning at 20% and Obama at 10%.

print this article

Return to marketing news headlines
View Marketing News Archive

Share with: