Don’t believe the US presidential opinion polls. Barring a political earthquake, Barack Obama will be re-elected at a canter
FROM tabloids and broadsheets to left-leaning blogs and conservative talk shows, the US media has been united on one point in recent months: the presidential election is too tight to call. The difference between the candidates is "razor thin", The New York Post said recently. The "race remains close", agreed The Washington Post. According to The New York Times it is "widely expected to rest on a final blitz of advertising and furious campaigning".
But it takes just a few clicks to go from that last article to one that tells a very different story - one much more in keeping with what science tells us about the election. The New York Times hosts FiveThirtyEight, a blog by statistician Nate Silver dedicated to crunching electoral numbers. It gives the Republican challenger Mitt Romney a 1-in-4 chance of victory. Over at PredictWise, another source of political forecasts, Romney's odds are only a shade better. The race isn't close or razor-thin or dependent on advertising. It is President Obama's to lose - something that readers are rarely told.
Why the discrepancy? To answer that question, think about what polls actually are. They are often taken as an indication of who will win the election. But polls only provide a snapshot, often with a large margin of error, of who would win if the election took place today. That's very different from what we really care about, which is the candidate most likely to win the real thing in November. That's a forecast. It's what FiveThirtyEight and PredictWise provide, and it's a more complex beast than a poll.
The PredictWise forecast, the work of Microsoft researcher David Rothschild, depends on three types of data and the impact that each is known to have.
One is economic indicators, and the link here is simple: the better the economy is doing, the greater the incumbent's chances of winning. The US economy remains unhealthy but, crucially, it's on the mend. We know from previous elections that the direction of the economy has a bigger influence than its absolute state, so this information narrowly favours Obama.
The next ingredient is the wisdom of the crowd. It is well known that groups can make more accurate predictions than individuals when opinions are aggregated into a collective forecast. In this case, the aggregation takes place at websites like the Iowa Electronic Markets, where investors buy and sell futures in the two candidates. The return on these contracts is based on who wins and by how much, so prices reflect the traders' collective confidence in each candidate. Obama's shares have recently been trading at two to three times the price of Romney's.
The final input takes us back to opinion polls. With a little number crunching, polls can be transformed into forecasts. The process depends on the trends in polling numbers seen during previous elections.
It's known, for instance, that support for the incumbent tends to pick up two to three months before election day. This is probably due to the challenger's honeymoon period coming to an end. It is easier for voters to idealise a challenger, who may not previously have had a high national profile, than it is the incumbent, who has been headline news for almost four years. As voters learn more about a challenger they inevitably discover things they do not like, prompting some to decide to stick with what they know. This may be one reason why Obama has edged ahead in the polls in recent weeks.
There are different recipes for combining these ingredients, and not every forecaster uses them in the same way. Some, including Rothschild, increase the accuracy by considering data on individual states rather than at a national level. But almost every model is predicting an Obama victory. Most have been making this prediction for a year or so. And election forecasting is hardly a newcomer: one model run by Allan Lichtman of the American University in Washington DC has correctly called the popular vote in the past seven elections. "I don't see how Obama can lose," Lichtman told US News & World Report.
If the models are robust, and their predictions strongly in favour of Obama, why are we being told that the race is a dead heat? I think it is partly a cultural issue. Earlier this year I wrote a story about election forecasting for a British publication. The science editor liked it, but a colleague on the politics desk vetoed the piece, in part because he simply didn't believe the forecasts. I can see why. The hurly-burly of day-to-day politics is filled with dramatic events, like the recent leaked video of Romney talking in unvarnished terms about voters he cannot hope to win over. These events make the race feel like a roller-coaster ride.
The truth, as revealed by the science, is much more prosaic. Obama is way ahead and has been for ages. The meat and drink of daily political reporting - party conventions, gaffes, attack ads - have a limited and often passing impact. That's not to say that an unforeseen event couldn't put Romney in the White House. But it would have to be something huge, because studies of previous elections show outcomes depend far more on fundamental factors such as employment rates.
That, however, doesn't make for an exciting story. I remember where I was when John McCain selected the deeply divisive Sarah Palin as his running mate in 2008 - an event that supposedly redefined the race. Two weeks later Lehman Brothers filed for bankruptcy. It was genuine drama, the stuff that sells newspapers and advertising space. But it didn't change the way people voted. The scientific predictions gave McCain minimal chance all the way through the 2008 race, and he duly lost. The forecasts are similar for Romney. The race is not tight, and the only honest approach is to say so.
Source:
http://www.newscientist.com/article/mg21528840.200-the-us-presidential-election-is-no-contest.html?full=true
Topic: USA politics Low Content Thread