US election polls latest 2016: What do the final surveys tell us about who will win?

Today’s US presidential election is a nightmare for the pollsters. On the one hand nearly every national poll puts Hillary Clinton ahead. Of some 20 or so final polls, all but two (one of which has long appeared to be an outlier) have put her in front. So anything other than a Clinton victory will be regarded as a polling failure.

But on the other hand, although there seems to have been a small swing in her favour in the final days, the Democrat nominee’s lead is still only a narrow one – on average between three and four points. Just four of the final polls give her a lead of more than four points.

That is not a lot more than the average error in presidential election polls during the last 50 years – two points. If the polls are overestimating her strength in this election by just a little more than that the outcome could be very close indeed.

Meanwhile, of course, it is not the nationwide vote that will decide who will become the next president. Rather, it will be who wins most votes in an electoral college in which big states have many more votes than small states and where, with just a couple of exceptions, the winner in a state takes all of its electoral college votes irrespective of the narrowness of their victory.

So much hangs on the accuracy of the individual state-by-state polls – especially those in the big states – and not on the nationwide polls.

These state polls underline the vulnerability of Clinton’s lead. The winning post is 270 electoral college votes. At the moment the polls only put her clearly ahead in states worth 248 votes. To get past the winning post, she needs to secure a couple of states – most likely Pennsylvania (20 college votes) and New Hampshire (4 votes), where her lead is a little less certain. At the same time, she has to avoid an upset in a state such as Michigan or Virginia that should be safely in her column.

Conversely, of course, she might yet win quite handsomely. If the polls are underestimating her lead by as much as two to three points, then another half a dozen states worth between them 81 votes could yet swing her way, enough to give her a 353 to 185 vote victory.

Certainly, if Donald Trump fails to pick up the big states that are currently on a knife-edge, most notably Florida and North Carolina, it is difficult to see how he will be able to defy the polls.

Doubtless at the back of Clinton’s mind will be the way Trump did sometimes defy the polls during the primary season. Meanwhile, the Republican standard bearer himself is fond of referring to the unexpected outcome (at least to some) of the EU referendum in Britain.

There are certainly some similarities between the pattern of Trump’s appeal and that for Brexit. In the UK, voters with fewer educational qualifications were more likely to vote to Leave the European Union. Similarly, Trump is ahead among those without a college degree – and well behind among those who do.

Four years ago, in contrast, those with and without a college degree divided between Barack Obama and Mitt Romney in much the same way. This change in the demography of the Republican vote creates a greater degree of uncertainty about the outcome – and, as in the Brexit referendum, places a premium on the polls having got the educational profile of those they interview right.

But in a contest as close as this one appears to be, much could depend on turnout. Forecasting who exactly will make it to the polls is always one of the hardest tasks facing pollsters. There will, in truth, be relief as much as triumph amongst the pollsters if, in the end, they have called the winner right.

John Curtice is Professor of Politics at the University of Strathclyde