David Blanchflower: Obama's win shows the numbers always count

Economic Outlook: We will teach about the failures of forecasters for years in statistics

So Obama won. It was one of the most bitter and negative campaigns ever, and living in a swing state we were bombarded with phone calls and attack ads. My family was divided; some strongly backed Obama, and others were Romney fans. Thankfully I left for the UK before election day, and was able to escape, and along with many I was an early voter and avoided the lines.

Before the results were known the pundits on the right such as Karl Rove and most Fox News commentators were convinced that Romney was going to win.

Peggy Noonan in The Wall Street Journal on 5 November argued: "I think it's Romney. I think he's stealing in 'like a thief with good tools', in Walker Percy's old words. While everyone is looking at the polls and the storm, Romney's slipping into the presidency. He's quietly rising, and he's been rising for a while … All the vibrations are right." He hadn't been rising actually. Bill Clinton's ex-pollster Dick Morris forecast a Republican "landslide". They thought there would be a huge Republican turnout because of the deep hatred for Obama. That didn't happen. Steve Forbes predicted on election night: "Mitt Romney will win big tonight."

We had our own clueless ones; Janet Daley, for example, in the Telegraph this week wrote "the election is – at this moment – so close that virtually all of the battleground states still defy prediction." No they didn't.

I was much struck by how uninformed the coverage was by the media in the UK who, right up until the last, said the race was too close to call, when it wasn't. The UK public was told that polls showed that the two presidential contenders were in a close race because the national polls suggested there was nothing to choose between them in the national voting. This just showed their ignorance, or perhaps was just designed to generate interest in the race. A US election isn't determined by who gets the most votes, but who wins the majority of the 538 votes in the electoral college. Each state has a set of electoral votes that are given in a first-past-the-post system where the winner takes all. The number of votes given is determined by the size of population. Big states like California and Texas with 55 and 38 votes respectively are much more important than little New Hampshire, where I live, which has four electoral votes.

The reality, though, is that the election was only ever about the so-called swing states. The country is so split that there was absolutely no chance on God's green earth that Romney was ever going to win California, Massachusetts, Connecticut, and the rest as Obama held double-digit leads. Conversely, Obama had no chance of winning Alaska, Arkansas, Texas, Utah, West Virginia, etc, as Romney held double-digit leads.

The two candidates barely campaigned or ran any adverts in these states, so any statistic that included them, such as the share of the popular vote, was utterly irrelevant as a predictor of the overall outcome and should have been ignored. The only places that mattered in the end were half a dozen states – Colorado, Iowa, Ohio, New Hampshire, Virginia, Florida. By election day Michigan, Nevada, New Mexico and Pennsylvania weren't even competitive, with big leads for Obama which were well outside the margins of error. South and North Carolina were going to be Romney wins, but they were too little and too late.

Over the final week four factors sealed it for Obama. First, the biggest weekly fall in oil prices for four years. Second, Hurricane Sandy, which Obama appeared to handle well. Third, New Jersey governor Chris Christie's praise of Obama's handling of the crisis; and finally, the pretty good job numbers that were announced just before election day.

This in many ways, though, was a victory for the pollsters, who called it right, and the pundits, especially the Republican-supporting ones, were left with egg on their faces. In particular Nate Silver, a number-crunching 34-year-old University of Chicago graduate at fivethirtyeight.com who tracked the polling data and called it spot on. By polling day Silver had Obama with a 91 per cent chance of winning, and called the races correctly in all 50 states plus DC. He did better than he did in 2008, when he called 49/50 states correctly. Other pollsters also did well; YouGov called it for Obama early, and the betting site Paddypower paid out a couple of days before the election. Intrade.com, which sold shares in the two candidates, consistently had Obama well ahead. It was never close. In all the polls that were conducted in Ohio, which was the crucial swing state, Romney was ahead in zero.

On election night Karl Rove even disputed Fox News' own call that Obama had won Ohio and hence was going to be President. In the end Obama won Ohio by 1.9 per cent. Another career bites the dust.

Of interest also was that a number of polling companies were hopelessly wrong. Rasmussen Reports, for example, had consistently showed a fall-off in Democratic party identification. Apparently a good part of Rasmussen's failures were because they sampled people by phone using only landlines. There is a growing trend among the young to use only cellphones, and Rasmussen simply missed them, which was a major problem given that the young disproportionately voted for Obama. Gallup's daily national tracking poll put Republican nominee Mitt Romney ahead by five points until it was suspended for Hurricane Sandy. We will be teaching about such supreme forecasting failures for years in statistics classes on how not to conduct surveys. Apparently most Republicans including the Romney campaign believed these flawed polls to the very last and thought they were going to win. The better modellers called it right. There are many lessons for the UK here, not least that it is hard to dislodge an incumbent even in a weak economy. But the biggest lesson is that spin won't beat the hard science.

Comments