Of all the controversies in a frenzied final week of a tense presidential election campaign, none has been more fraught than the small matter of who is most likely to win.
One view, shared by statistical forecasters, betting markets and state-level opinion polls, has Barack Obama as a strong favourite — perhaps something more than that. The highest profile of the forecasters, FiveThirtyEight wunderkind Nate Silver, gives Obama an 86.3% chance of victory at the time of writing. An even more bullish view of Obama’s prospects is offered by Silver’s equally credible rival Sam Wang, a neurologist who has been putting his powers of analysis to work for the Princeton Election Consortium. Wang has two models on offer, one of which puts Obama’s chances at 98.3%, the other at 99.9%.
Feeding into these forecasts have been state polls showing Obama with small but stubborn leads where he needs them most. Feeding out of them has been an expectation of an Obama victory in betting markets, albeit in less comprehensive terms (the widely quoted Intrade has it at 66.7%).
The alternative view — that Mitt Romney is at level pegging or better — has been advanced by right-wing bloviators, media outlets with a ratings-driven interest in a close contest, journalists of the old school who dislike the way their game is changing, and — the one persistent basis for real doubt — polling conducted at the national rather than the local level, some of it of excellent pedigree.
Whereas the Australian political textbook instructs leaders and candidates to claim underdog status, presidential campaigns are concerned with generating enthusiasm among supporters in order to drive turnout. It’s thought this is best achieved by projecting an image of strength — perhaps especially so in the case of the Republicans, who are pitching to a male-dominated audience. Servants of the Republican cause have thus been eager to discern in the polling and forecasting professions the same liberal biases they feel they have come to know from scientific organisations and the media.
Forecasters have also come under fire from establishment media figures who argue political insight can be gained only through insider access, such as is available only to members of the journalistic closed shop. Leading the charge last week was cable news host and former Republican congressman Joe Scarborough, who thundered that the level of concern he was hearing from the Democratic camp was risibly inconsistent with the one-in-six chance of defeat being projected for them by Silver (evidently Joe hasn’t played Russian roulette too often).
More troubling from Obama’s perspective has been the national polling, the RealClearPolitics aggregate of which had Romney about a point ahead from the aftermath of his October 3 debate win until Hurricane Sandy hit on October 29. The worst results of all for Obama came from the most prestigious name in the business, Gallup, which consistently had Romney’s lead at about five points up until its tracking poll closed shop after Sandy hit (today it has returned to the field with a final poll that has Romney only one point in front).
Confoundingly, polls conducted at state level — in many cases by the same organisations that were conducting the national polls — showed a weight of support for Obama that was incompatible with the national figures. This has most famously been reflected in a barrage of polling putting Obama 2-3% ahead in the key battleground of Ohio. Polling aggregates have also had him ahead in the swing states of Colorado, Nevada, New Hampshire and Iowa — enough to get him over the line without assistance from Virginia, Florida or North Carolina, where he has been at least competitive.
National polling conducted in the wake of Sandy has found Obama edging upwards, narrowing the state polling gap and further weakening the case that Romney has as much grounds for confidence as his campaign asserts. Nate Silver’s state polling-inclusive model now has Obama 2.1% ahead on its national vote projection, against an averaged 1.1% on the basis of Sunday’s national polls. Even the national figure taken in isolation would give Romney little hope of victory, except at the outer reaches of the error margin.
Certainly it is not beyond the realms of possibility that systemic inaccuracy is causing the pollsters to get it wrong. Polls do better at some elections than others; 1980 and 1992 are examples where the combined error was of such size as would put the current result in doubt (although it might be thought notable that there were substantial campaigns by independent candidates on both occasions).
Such is the level of subjectivity involved in polling a US election, particularly in modelling turnout, that there is always a reasonable basis to argue that the wrong assumptions are being made. Romney boosters point to his lead among voters who identify as independent, and argue the pollsters are wrong to credit Obama with the level of Democratic turnout needed to cancel it out. However, party identification is at all times a slippery concept, and it has been complicated over the past term by the rise of the Tea Party and its adherents’ conviction that they are independents rather than Republicans.
The pollsters’ and forecasters’ judgements on such matters may well be imperfect, but they are unlikely to have been motivated on any level by wishful thinking, which clearly can’t be said for most of those talking up Romney. For this reason, I don’t see any reason to bet against the view shared by FiveThirtyEight and the RealClearPolitics state polling averages: that Obama will win the electoral college 303 to 235, carrying Ohio, Virginia, New Hampshire, Pennsylvania, Iowa, Colorado and Nevada, while falling short in Florida and North Carolina.