Clinton clings to tiny lead — but 'Bradley' factor could be skewing the polls
Is Donald Trump's support being underestimated by pollsters?
Two factors indicate it's possible.
1. The Brexit effect
UKIP leader Nigel Farage says lots of people who don't usually vote will cast a ballot for Mr Trump, mirroring the so-called "Brexit effect" in the UK that caught pollsters off guard.
There's a possibility this is happening. As I type, just over 40 million early votes have been cast [UPDATE: the final early voting tally was 47 million].
Early votes have been coming in at around 1.5 million to 2 million a day, so it would take quite a last-minute surge to beat the 46 million early votes in 2012 (of a total 128 million).
Yet in some key areas, earlier voting has been heavier. Exhibit A is that battleground-of-battlegrounds, Florida, where early voting closes today NZT (most states will keep early voting open).
Figures post this morning show 6.1 million early votes cast, compared to 4.7 million in 2012.
So the Brexit effect could be in play.
Against this, the Clinton campaign has a far larger get-out-the-vote field operation, and surveys indicate that Hispanic voters are turning out in greater numbers than 2012 — unlikely to be a positive for Mr Trump.
And in key swings states, including Nevada, registered Democrat turnout is up over 2012. But the lingering question is: Are they staying loyal? How many white working class supporters are crossing over, a la the blue-collar "Reagan Democrats" in 1984?
The money race (Source: WaPo). Clinton's campaign has spent 93% of its money raised, Trump's 96%
2. The Bradley effect
In the 1982 race for Governor of California, black candidate Tom Bradley (a Democrat), enjoyed a lead in the polls but lost to white Republican George Deukmejian.
Poll historians call this the "Bradley effect." Some people lied and told pollsters they would support Mr Bradley because they did not want to appear racist.
Variations on this theme include the "Shy Tory" effect in the UK in 1992 when some people were too sheepish to tell pollsters they would vote for the Conservative Party as led by the unfashionable John Major (his party was behind in the polls, but won). And the "Wilder" effect 1989, where polls showed Democrat candidate Douglas Wilder comfortably on track to become Virgina's first black governor — but ultimately he only won his race against his white Republican rival by a razor-thin margin.
Are some Trump supporters also too sheepish to declare their support?
This cycle, due to the spiralling cost of trying to reach people by live phone as due to so many ditching landlines, around half the polls are online (with online panels weighted to match census data).
Pundits say voters are more likely to express their true preference with an online form rather than when they talk to a human pollster.
Morning Consult, which has been conducting a tracking poll for Politico, decided to test the "Shy Trump voter" hypothesis by conducting phone and online interviews with a sample of 2075 likely voters (read its full report here).
The test found there is a "social desirability" effect, which is quite marked among higher income and college educated voters (blue = Clinton, red = Trump).
But once voters across the board are factored in, the "Shy Trump voter" effect is a lot smaller; around 2%:
Politico deems that 2% too small to influence the race ... but bear in mind we're now talking about a race where Clinton has a margin of two points or under according to the latest poll-of-poll surveys.
Some real-life evidence runs against it. Some online surveys, such as that conducted by IPSOS/Reuters, which has Clinton in the lead by four points as of this morning, are actually more bullish for the Democrat that some live phone surveys, such as the one conducted by the Trump-friendly Fox News, which as of today actually shows a tighter race with more people avowing support for Trump to put him within 2 points of the lead.
You could also argue for some degree of a "Shy Clinton supporter" effect, given the Democrat's flat campaign and various baggage.
But overall, the 2% "Shy Trumper" effect vs a 2% Clinton lead, and a possible "Brexit" effect boosting Trump's vote mean this election is too close to call.
The Democrat's best hope remains that the race comes down to the state-by-state electoral college vote, where she maintains a narrow lead in a couple of key battlegrounds that make it tricky for Trump to get to the magic 270 needed to take the Whitehouse (Politico has a good summary here). But even in her so-called "firewall" states, Clinton's lead is still close to the margin of error.
Nigel Farage might be deeply unloveable, but his Brexit poll theory proved correct — unlike polling guru Nate Silver at FiveThirtyEight, who called it wrong.