With barely 24 hours to go until polls open in the US Presidential Election, I thought I’d take a look at the chances of each candidate. I’m going to try to put personal prejudices to one side and look at this impartially – as a former Maths teacher, it seems appropriate to simply look at the evidence.
As brief background, for those who don’t know how this all works, the American Presidential Election is decided not based on who wins the most votes but on the ‘Electoral College’ system. Each State has a certain number of votes in the Electoral College based on its population; California has 55 compared with New Hampshire’s 4, for example. There are 538 Electoral College votes in total, so a candidate needs 270 to win. A 269-269 tie, or a three-candidate race where nobody gets 270, leads to Congress choosing the President and vice-President according to arcane rules. Within most (not quite all) States, it’s winner-takes-all so ‘winning’ a State grants the victor all of that State’s votes. The whole thing’s a little more complicated than that, but I’m sacrificing accuracy for brevity. What this all means is that the winner of the popular vote usually becomes President, but not always. And nobody will ever really mention states like California because it’s a safe state which can be relied upon to vote for the Democrats. Like marginal constituencies in the UK, all the resources are poured into the marginal (‘swing’) states of the USA.
I think it’s fairly clear that a couple of weeks ago, Clinton was in the lead. The decision by the FBI to re-open the investigation into her coincided with a clear narrowing of the gap between her and Donald Trump. The gap had been shrinking slightly for some time, though, so it’s quite possible that this wasn’t the only cause.
1) Comparisons between Trump and Brexit polls are not a good argument for a Trump victory. The media narrative tells us that Brexit happened despite the opinion polls, but that’s just not true. 16 of the polls conducted in June showed Leave ahead, 16 showed Remain ahead, and 1 showed a tie. The notion that Brexit was losing was based on media groupthink. It assumed that the online polls were wrong and that phone polls were correct. The headline figures in the UK polls were downweighting Leave, so the raw data would often have Leave ahead even if the published poll showed Remain ahead. American polls occasionally do this, but far less often than British ones.
2) Postal voting helped Brexit, but early voting won’t help Trump. A week before the British referendum, one poll showed Leave in the lead by 10%. A swing to the status quo in the final week showed that some voters were falling for the rhetoric and propaganda in the last week. But many votes were already in the bag. Leave won the postal votes by a huge margin. It would have taken a landslide Remain vote on the day for the referendum to go to Remain. Trump’s problem here is twofold: firstly, when coming from behind, early voting will hinder him as much as it helped Brexit. Secondly, in states with published early voting figures, it seems that Democrats have been more likely to turn out than Republicans. Trump could, like Remain in the referendum, need a landslide on the day.
3) The Electoral College system probably favours Trump. On average, Donald Trump is around 3% behind in the national polls. But that’s not anything like as bad as it seems for Trump. If he loses the popular vote by anything under (say) 1.5%, he could still win the election by carrying enough states to take the Electoral College.
4) Third-party votes increase the likelihood of a polling error. Americans rarely vote outside the two-party system. Yet the polls are finding a Libertarian candidate, Gary Johnson, on roughly 5% of the vote nationally. A Green candidate is also showing up in most polls. One of the reasons that the UK is so difficult to poll is that there are so many parties: the ‘big three’ of Conservatives, Labour, UKIP (I couldn’t resist) – plus the Liberal Democrats, Greens, SNP and Plaid Cymru (Northern Ireland is usually polled separately) who all have at least localised or regional support. Opinion polls often don’t pick up tactical voting. If so, could some Johnson or Stein voters move back to the better-known candidates as a last-minute tactical vote?
5) If the ‘shy Trump’ factor is real, it could swing the election. I’ve never been totally sure that I buy this one. Clinton isn’t exactly a popular Democrat. Many people won’t want to admit to voting for Clinton either, especially after the latest scandal. And Trump supporters are incredibly vocal. I could see this one working either way, but the uncertainty is probably still a good thing for Trump. If you’re slightly behind in the polls, you want uncertainty. If you’re slightly ahead, you want stability – and Clinton doesn’t have that.
6) Trump’s Utah problem may have gone away. There were a few polls in Utah showing independent candidate Evan McMullin in a close second place, or even winning, in Utah – where Trump hasn’t gone down well with a conservative, religious vote. That threat seems to have subsided slightly (though there’s a caveat that volatile races are more difficult to poll). If Trump lost those 6 Electoral College votes in a deeply Republican state, it would make his path to the 270 Electoral College votes needed to win that much harder. The apparent move away from McMullin is good news for Trump.
7) Nevada, though, is a growing problem for Trump. In an election-winning year, the Republicans really should be winning Nevada. But in 2012 Obama carried it by 7 points over Romney. That was thanks, mainly, to a large lead in early voting for Obama. This year, it appears that Clinton has an even larger early-voting edge than Obama did. If Trump fails to gain Nevada, he’ll need more breakthroughs elsewhere.
8) The same applies to Florida, but to a lesser extent. I can certainly see Trump winning Florida, though it’ll be a tight race. Florida is a big early voting state, which could hurt Trump. The difference between Nevada and Florida is that Trump has a much smaller mountain to climb in Florida. Likewise North Carolina. It’s just about conceivable that if the early voting has made such a huge difference, Florida, North Carolina and Nevada could all vote Democrat and that Clinton could win even if she loses the popular vote. This, in my opinion, is far less likely than Trump winning despite losing the popular vote.
9) There are big demographic changes since 2012. Hispanic voters are far more enthused about Clinton than they were about Obama; the reverse is true about African American voters. Middle and upper-class white voters seem to be deserting Trump, whereas working-class white voters are much more likely to vote for Trump than they were for Romney. The gender divide is stark: you may have seen the news that one poll only had the same male/female split as in 2012. Well yes, but there’s a reason why that made news: pretty much every other poll has showed that the gender split is even bigger than it was then. Women will favour the Democrats, and men will favour the Republicans – probably by an even greater margin than last time.
10) This means that turnout really matters. Here’s where comparisons with Brexit do work: if we hadn’t got working-class voters out to the polls in such numbers, we’d have lost the referendum. Given that this demographic tends to vote less often, we had to succeed in that regard. Trump has the same issue: he must get his supporters out. This is where polls can often make mistakes; they often assume those who haven’t voted recently won’t vote again this time either. That may be a false assumption with regard to Trump, and could lead to polls underestimating his support.
11) How does Trump get to 270 Electoral College votes? If we assume that Trump wins Florida, North Carolina, Arizona, Iowa and Ohio, he’s still behind in the Electoral College. I seriously doubt that he’ll win Nevada for the reasons above. But even then, he also needs one more state. New Hampshire might (just) do it, leaving the race 270-268 or even a 269-269 tie (depending on how Maine’s second Congressional District votes – Maine, unusually, isn’t a winner-take-all state). Trump has a very tenuous route there to eke out a tiny victory, but I would look at two other states which haven’t been polled that much as potential pointers to a hypothetical Trump win.
12) Michigan and Wisconsin are states to watch out for. If Trump does well in one of these states, he’ll probably do well in the other because they’re very demographically similar. Win Michigan, and he can cope with losing Nevada (which is my stick-my-neck-out prediction: that’s the one State where Trump’s generally ahead in the polls that I’m pretty convinced he will lose). Win Wisconsin, and we’re back to the 270-268 or 269-269 scenarios above. Win them both, and suddenly Clinton would need to win Florida or Ohio. I mention Michigan and Wisconsin rather than Pennsylvania or Colorado partly because of demographics, but partly because Michigan and Wisconsin have been polled less often so an error becomes more likely. And Colorado, say, may not make that much difference in isolation: win Colorado but lose Nevada, and Trump is still behind. Even winning New Hampshire, which is on a knife-edge, wouldn’t help him then.
13) Back to the key point though – a small Trump surge would likely be enough to make these things happen. Clinton has a problem in that her best demographics are under-represented in the swing states. It’s one of the reasons why this year it’s highly likely that she’ll lose the election even if she wins the popular vote by up to 1.5%.
14) Could Clinton benefit from a late swingback? If the FBI re-opening its investigation into Clinton had cost her some votes, she may regain them now that they’ve said there’s no new evidence to justify charging her with a criminal offence. But if so, the movement is unlikely to be picked up by the polls. Such news takes time to filter through to the general public, and polls take time for fieldwork to be conducted. When the news broke, just 36 hours before Election Day, it’s unlikely that any impact will be seen.
The bottom line? This is an unusual election. Two weeks ago, the data suggested that it was unwinnable for Trump. Today, things have changed. The Clinton average 7% lead would have carried her to the White House, I’m sure. An average Clinton 3% lead is flaky. It includes some polls with a tiny Trump lead, ranging all the way to polls suggesting an easy Clinton victory. She has pretty much zero margin for error.
Finally, treat individual polls with caution. ‘One in every 20 polls is a rogue poll’ (I’m not explaining that in detail, to avoid statistical tedium) even before accounting for methodology issues. Polls in the UK generally have larger sample sizes than those in the USA, so the margin of error is even wider. If you see a single poll with Trump winning, or a single poll with a massive Clinton lead, check other polling companies’ results. Don’t cherry-pick the one that shows what you want to see.
Overall, behind all the detail, the message is pretty simple. If the polls are right on average, or under-representing Clinton at all, then she wins. If they’re under-representing Trump even a little, then we could well be seeing him in the White House.