Washington, November 8 (The Guardian): The eleventh hour prediction in the US Presidential election is that Hillary Clinton’s support is at 46.8% while Donald Trump’s is at 44.3%.
When analysts refer to the Democrat’s 2.5 percentage point lead, they are talking about the difference between those two figures in polling averages.
It’s an important number but it’s also probably an inaccurate one. That’s because polling averages do not capture fully a few factors that will affect the final outcome.
More than 41 million Americans have already cast their votes. Some, but not all, polls include such early voters. More than a third of respondents in a survey released by Bloomberg on Monday morning said they had already voted. The poll showed Clinton leading by three points.
There are some other positive signs for Clinton. Turnout of Hispanic voters is up in every state where early voting data is available – and polling data suggests that as the weeks have passed, those in this group of Americans have increasingly said they do plan on voting. Since three-quarters of Hispanic Americans say they will vote for Clinton, that is bad news for Republicans.
What’s more complicated, though, is figuring out how increases in early turnout among Hispanic voters might offset declines in early turnout among African Americans and voters aged 18 to 29. And it’s also possible that voter behavior differs by political affiliation. It might simply be that those who support Trump are more inclined to show up on election day, rather than voting ahead of time.
In Nevada, where polls suggested Trump was leading by 1.5 points, early votes look positive for a Clinton win. More than half of active votes have been cast, 42% of them among registered Democrats and 36% among registered Republicans.
To win Nevada, Trump would have to win over a majority of independents, which would mean bucking trends from 2008 and 2012. Which is possible. But maybe Trump has already written it off – with only six electoral college votes available, it might not be crucial in getting either candidate to the 270 they need to win.
Historically the popular vote (the results as a percentage of all ballots cast) doesn’t neatly translate to the electoral vote (the results as a share of 538 electoral college votes). Four times, most recently in 2000 between George Bush and Al Gore, the winning candidate has won the electoral vote but not the popular vote.
That affects the way we read the polls. Having 47% support in national polls does not mean Clinton is also set to win 47% of the electoral college votes. The size of her support varies considerably from state to state. In Maryland, she is polling at 63%. In Idaho, she has around 23%.
So, any definition of a battleground state needs to take into account two things: how close the two candidates are in the polls and how many electoral college votes are available in that state. If you set your criteria as being 1) that the candidates are separated by six points or less in state polls and 2) that the state is worth 10 electoral college votes or more, the battleground states are as follows: Florida, Pennsylvania, Ohio, Michigan, Georgia, North Carolina, Virginia, Arizona, Wisconsin and Minnesota.
Based on early voting and the fact that nearly one in five eligible voters there is Hispanic, Florida looks likely to vote for Clinton. That’s important, given that the state is worth 29 electoral college votes – but it might not be enough. Even if Clinton can add Florida to the list of states she can count on, that will only take her to 216 of the 270 electoral college votes she needs.
The margin of error is crucial. Typically, when a poll is published, buried in the methodology section is a line that says something like “margin of error ± 3.5 percentage points”, which means for example that Clinton’s actual level of support could be as high as 50.3% or as low as 43.3%.

Since the two candidates are so close in the polls, that margin of error could be the difference between winning the popular vote or losing it.
If the margin of error systematically favors the frontrunner or the underdog, we could just take that into account. Unfortunately, it doesn’t. Take Gallup poll, for example, one of the country’s largest polling firms. In 2012, it overestimated Barack Obama’s vote share by two points. In 2008, it underestimated it by two points.
The polling average simply takes all the credible polls conducted at a similar time and averages out their results. It rarely takes into consideration the past accuracy of polling companies, let alone how many people were included in the poll and whether pollsters spoke to enough minority groups. Polls are very different.
NBC News suggests Clinton is ahead by seven points while the Los Angeles Times poll shows Trump leading nationally by five points. Encouragingly for Clinton, survey companies regarded as more reliable are pointing to a larger lead for her.
It’s unusual to have so much disagreement so close to election day. But this isn’t a typical election: it is one that has been characterized by uncertainty and surprise.
What we can say with absolute certainty based on the polling data is that a lot of Americans are going to wake up disappointed on Wednesday.