Close election? If you ignore the data

The ‘close’ US election is a media invention. If Obama loses, we might as well give up on polling data altogether.


For a columnist to make a prediction on the eve of an election is a fool’s errand. The warm glow of confident smugness lasts less than a day, and if it’s wrong his idiocy remains on show in perpetuity.

Nonetheless, I’m going to predict Democrat Barack Obama will be the US president elect again by this time tomorrow, having won the electoral college vote 303-235.

It’s not much of a prediction. I’ve simply assumed each state’s result will follow the average of the most recent opinion polls. In fact, I didn’t even need to work it out for myself. Websites such as RealClearPolitics and the New York Times’ FiveThirtyEight blog have done it for me.

Depending on the methodology, the predicted state-by-state results vary slightly, but the consensus is an Obama win. FiveThirtyEight, run by the equally celebrated and villified statistician Nate Silver, actually gives Obama an even wider victory margin than I do, and a staggering 92 per cent chance of winning overall.

Yet despite the handy, freely available number crunching, the 2012 race to the White House has constantly been called “neck and neck”, a “coin toss” and “too close to call” in media reports. Well, if commentators believe that, they must have no faith whatsoever in the polling data they spend so much time reporting.

In reality, they don’t believe their own story. Opinion polls were pointing to an Obama win well before Mitt Romney officially became the Republican presidential nominee in August. The trend has barely altered since. Even after the first TV debate between the two men, which Obama was generally agreed to have lost, he never lost the lead in the swing states that determine the result.

There are several reasons that the news industry, both in the US and internationally, has been less than candid; chief among them being the need to create a nail-biting narrative that keeps political junkies hungry for more – and which at the same time boosts their readership and viewing figures. Then there’s the US media’s compulsive nervousness about reinforcing accusations of a liberal slant by giving Obama the advantage.

Polls can be unreliable, of course. They vastly overestimated the Liberal Democrats’ share of the vote in the 2010 UK general election. They’ve previously given phantom ‘wins’ to US presidential candidates such as George McGovern in 1972 and Thomas Dewey in 1948, both of whom went on to be defeated. The Chicago Tribune newspaper even went to press on election night with the front-page headline “Dewey defeats Truman”, much to its embarrassment the next day.

But the variety of polling surveys and methods available in the US now is unmatched, and on the swing state averages, Obama has been ahead – if only by narrow margins – for virtually the entire race.

Even if you ignore the long-term trends and automatically award Romney the win in all the swing states where he’s ahead in the polling average today, or where he trails Obama by no more than two percentage points – New Hampshire, Virginia, Florida, Colorado and North Carolina – he still loses.

Of course, there’s a chance the polls are all wrong. There’s a chance they’ve all been wrong all year. But if that’s the case, we need to make sure we never look at one again, and instead judge every political contest by the the strength of media bluster blowing one way or the other.

Tables at Marketing Week’s Data Strategy awards are now available for booking. To secure seats for you and your team at the premier event to celebrate data-driven marketing, click here now.



    Leave a comment