14 November 2016

How did Trump win?

Yep, this is another one of my stats pieces, so if you're looking for lots of speculation about sociological factors, you're not in the right place.

Donald Trump defied the odds to win the US Presidential election. But how did he defy the odds and the polls? What did he do that we got completely wrong? Let's delve into it to find out.


Trump really had no right to go down the board as far as he did, really. Taking Michigan, for example: Trump needed a 4.8% swing to take this and he got a 4.9% swing here. We don't have any figures yet for the congressional districts in Maine and Nebraska, but all throughout the night the the BBC were predicting that the second one in both states would swing to the other party. I found this odd considering how deep they were in "safe" territory on both sides. As it was, Nebraska II stayed red.

Whilst 306 vs 232 seems like a big margin, it really wasn't. Had Michigan, Pennsylvania, and Maine II gone blue, we would have had a 269 vs 269 tie, and then, would Paul Ryan really have voted for Donald Trump in the House of Representatives? We will never know. 

Just how uniform is America? Actually, all the swings in each particular state were close to the national swing of 1.9% from Democrat to Republican. We can further look at this by use of a box plot, indicating the swings in each state. 


Immediately, your eyes are drawn to Utah, where there was a huge swing to the Democrats. There wasn't, really - Evan McMullin ran as an independent and gained real momentum, taking many votes off Donald Trump, but ultimately unsuccessful in his bid to win the state. And that leads me onto the third party effects.

Gary Johnson hit 9% in New Mexico but it really was a terrible election night for him, and he is the reason Donald Trump won, one might argue. Two days before election day, Johnson was polling at around 7%, and with Trump looking like he couldn't win, Johnson's campaign team released a video telling the public that a vote for the Libertarians would be a vote to block Hillary Clinton from becoming President, as Donald Trump couldn't win...


As I explained in my first blog post on the election, Johnson's support was mainly from dissident Republicans who had voted for Mitt Romney in 2012 but did not want to vote for Donald Trump. But it appears they did not recognise this fact; throughout the campaign Johnson's campaign targeted dissident Bernie Sanders supporters, highlighting his similarities on social and foreign policy to Sanders'. But by telling his supporters that Trump couldn't win, in effect he was telling them it was safe to vote for Trump. Oops. On election day, his poll numbers dropped by 3% and Donald Trump's increased by 3%. Gary Johnson's campaign manager put Donald Trump in the White House.

But let's go back to the swings between the Democrats and the Republicans, The national swing, as has already been stated, was 1.9% from Democrat to Republican. Not enough on a uniform national swing? Correct. But in the swing states, that figure was much higher, 2.8%. It is therefore a great targeting strategy from the Trump campaign, and poor from Clinton's campaign that she did not visit Wisconsin once. Wisconsin should have stayed blue; it went red.


As I've already said, the biggest Democrat swing was in Utah, but aside from that only 10 out of 51 "states" for which we have data (we are missing the 5 congressional districts, but we do have the District of Columbia) produced swings from Romney to Clinton. The other 40 all produced some sort of swing from Obama to Trump.

Clinton, to give her credit, did very well to hang on to Virginia and Colorado. At the beginning of the night, Arizona and Georgia were not "called" immediately, and as these were safe in Republican territory, this was a good sign for Clinton. Nonetheless, the fact that these states stayed red despite swings to Clinton showed us just how terrible the strategy was - that they decided to attack marginal Republican states rather than defend their own states like Wisconsin and Michigan. Her attack board was a complete failure:


Trump missed Virginia and Colorado (number 3 and 4 targets), but he picked up states lower down the board that required such a large swing we didn't even consider them individually - Michigan, Maine II, and Wisconsin. This compensated for his failure earlier on.


For the Democrats to win the presidency back in 2020, the easiest path to victory (notwithstanding the congressional districts) is to take Michigan on a 0.2% swing, Wisconsin on a 0.5% swing, and Pennsylvania on a 0.6% swing. That would put the Democrats back in the White House - and it gives Donald Trump no room to mess up. At all. Ohio may be out of reach for the Democrats (4.3% swing required), but it may not be necessary.

2 November 2016

US Election 2016 - Update

There's now under a week to go until election night, and since my blog post in May, a lot has happened, but this is what the ramifications mean.

As I said in May, it's incredibly difficult for the Republican candidate to win, irrespective of the fact that person is Donald Trump. As such, on a uniform national swing, he needs a 2.7% swing from the Democrats to the Republicans to win. In other words, everyone who voted for Mitt Romney in 2012 needs to vote for Donald Trump, as well as 2.7 out of every 100 people who voted for Barack Obama.

A poll of polls is the cumulative amalgamation of several polls over a very short period. So I'm going to construct my own poll of polls, looking solely at polls published on 1 November. They show the following:


This was weighted in favour of polls with larger sample sizes. Now, what does that mean on the change on 2012, and then on the Swingometer?

On a Uniform National Swing, Donald Trump would become president as things stand. The biggest problem for him, however, is that he's not, in terms of constituency state polls, doing enough. He wouldn't win Virgina (#3 on his target list), Colorado (#4) and Pennsylvania (#5), Incredibly, that would be 269 vs 269. Otherwise, the swingometer is consistent uniformly. Making a bad situation worse is that there would be a Democrat gain in North Carolina. Clinton is on her way to the White House.

Furthermore, the indepedent Evan McMullin in Utah means that Donald Trump could lose 6 electoral votes.

A lot of air has gone out of Gary Johnson's balloon sadly, meaning that his potential targets in New Mexico, for example, are heading to Clinton and not to Trump. Johnson's highest state at the moment is Maine, but he's 34 points behind Hillary Clinton there. Gary Johnson can stop Hillary Clinton from becoming President. Donald Trump will not win.