Predicting the US election

[aasim aslam abbasi]

For a few days, just before and immediately after Hurricane Sandy caused unprecedented devastation along some of the most heavily populated areas of the US Eastern seaboard, the upcoming, hotly contested US elections almost disappeared from American television media.

Many Americans heaved a sigh of relief to be rid of the incessant and increasingly vitriolic politicking, even if it were because of tragic reasons. US news media has been hammering away about these elections for most of this year, in usually unabashedly partisan ways (but more on that some other time). In the wake of Sandy, this minute dissection of electoral politics was replaced with wall-to-wall coverage of the superstorm’s after-effects. However, with only one day to go till America picks a new president and elects half its parliament, politics has slowly but surely begun to take centre stage in the media again.

America is a country obsessed with trying to see into the future. And there is no greater industry at the moment than trying to predict, through opinion polls, whether Democrat President Barack Hussein Obama will retain his job or whether he will be ousted by his Republican challenger Willard Mitt Romney. For an outsider — and one suspects for most Americans — the sheer number and diversity of polls, administered by a slew of competing media organisations and private think tanks, is bewildering. It doesn’t help that some of them seem to directly contradict each other.

For example, while Gallup’s poll conducted just before Sandy hit put Romney ahead of Obama by a healthy margin of 51 per cent to 46 per cent, the RAND Corporation’s poll at the same time had Obama ahead by an even bigger margin (51 per cent to 45 per cent). Most other national-level polls, however, seem to show both the incumbent president and his rival in a statistical dead-heat, with the one or two point leads either way being within the polling margins of error.

But polls have become increasingly important in America as instant bellwethers for changing public opinion. It is important to understand that, in general, US polling methodology has progressively become fairly robust, especially over the last two decades, and its results have, unlike polling data in countries such as Pakistan, proved to be mostly an accurate representation of the actual electoral mood of the country. This is the reason why both major parties are also obsessed with the data and have their spokespersons spend countless hours on the media trying to either crow over their results or dismiss negative ones and ‘spin’ an analysis to their advantage. Each day it seems brings multiple little election results before the elections have even taken place.

But aside from showcasing that America is a deeply polarised country — along ideological or party lines (as the dead heat of the national popular vote indicates), along gender lines (more women support Obama), along race lines (Romney has a wide lead among white men while more than 90 per cent of blacks and a sizeable majority of Latinos support Obama) and along demographic lines (the younger tend to favour Obama, the older Romney) – are these polls actually painting an accurate picture of who will win come Nov 6?

Not so if you ask Nate Silver, blogger-statistician turned celebrity political pundit, whose independent blog titled ‘FiveThirtyEight’ – named after the number of electors in the US presidential electoral college – was bought out in 2010 by The New York Times after only two years of operation and now is a prominent feature on the Times’ own blog site. Much to the discomfort of the Republicans, Silver has been arguing for quite some time that not only is President Obama overwhelmingly favoured to win the coming elections – he now gives Obama more than a 85 per cent chance of winning – he may well do so with something akin to a landslide.

At the heart of Silver’s analysis of aggregated polling data are two elements. The first of these has to do with the idea of the electoral college – similar to how Pakistan’s president is elected in a sense — where 538 electors rather than the popular vote will determine who becomes the next president. While national-level polls might tell you how the popular vote may fall, they are not very useful in predicting a winner in a first-past the post system, where a win by even one vote in a state can mean all the electoral votes of that state go to the leader. So the state-level polls matter far more to him than the national polls and particularly in the so-called ‘swing’ states which could tilt either to the Democrats or Republicans.

The second element is Silver’s ‘secret sauce’ so to speak, where he ‘weights’ others’ polls according to their pollsters’ past accuracy as well as other factors such as demographics and voting patterns. And according to his current analysis, the probabilities simply do not favour Romney. Silver is not an easy statistics-geek to dismiss: in 2008 he correctly predicted all of Obama’s electoral college wins except in Indiana and all the results of the Senate races.

As political commentator Michael Tomasky explained in a piece in early August in The Daily Beast — another blog, affiliated with Newsweek — Silver’s current analysis may not be liked by the Republicans (who are championing conservative analyst Michael Barone who predicts a landslide for Romney), but even Democrats do not want to talk too much about it, lest they jinx their chances with complacency. More interestingly, Tomasky argues that the television media has by-and-large down-played Silver’s predictions because they would undermine the narrative of a close election, and thus their ratings.

It’s all a far cry from considerations of who will better manage the struggling American economy, be better on divisive social issues, which candidate seems more ‘presidential’ and which party will actually manage to get its members out to vote — as the media keeps saying will determine these elections.

Nate Silver has not actually said that Mitt Romney cannot win under any circumstances, he is simply giving him only a 15 per cent chance. But if Romney is actually able to pull off such an upset or Barone’s predictions come true, it won’t just be Obama out of a job.

Source: daily dawn
http://dawn.com/2012/11/05/predicting-the-us-elections/